protonscr

The Witcher 3: Wild Hunt — Remastered

protonopen appid 292030Game compatibility - UnofficialRegressionXAudio2
ValveSoftware/Proton#607 · opened 2018-08-26 by StructByLightning · updated 2026-10-13 · 642 comments · github · game page · search this game
SStructByLightning 2018-08-26 github
  1. Rotfiends are invisible. Still present with Nvidia 396.
  2. The Wild Hunt hounds (battle in the elven ruins with Keira Metz) are invisible.
  3. When in bordered window mode (1600x900), alt-tabbing out of the window works. However, the mouse cursor occasionally reverts back to the Witcher cursor if the Witcher window is underneath the foremost window. When it does this, the Witcher window does not have focus and I can't click on the foremost window. Still present with Nvidia 396.
  4. Clicking to focus on the Witcher window works, but the cursor is no longer locked to the window. Minimizing the Witcher window and then bringing it back up temporarily solves the issue. Still present with Nvidia 396. However, if you click the black menu (but not on a button) when clicking to focus the Witcher window, the cursor will become locked again once you resume the game. Super weird. I don't think it worked that way before. The menu fix doesn't always work either.
  5. Audio occasionally becomes staticky and distorted. Restarting the game fixes the issue.
  6. Some buttons have weird hit boxes. For example, the slider to bet during a Gwent game requires you to click below the graphic. Still present with Nvidia 396.
Kkisak-valve maintainer 2018-08-26 github

Hello @StructByLightning, please copy your system information from steam (Steam -> Help -> System Information) and put it in a gist, then include a link to the gist in this issue report. Also, are you testing the GOTY edition?

SStructByLightning 2018-08-26 github

Link. I don't think I'm using the GOTY version. I think I bought the base game first and then the expansion pass, but I don't remember for sure. Is there a way to tell? In my library it's titled "The Witcher 3: Wild Hunt".

Kkisak-valve maintainer 2018-08-26 github

You may see better behavior with nvidia 396.51 or newer.

SStructByLightning 2018-08-26 github

Updated and you're right. Performance is much better, I'm staying locked at 60FPS nearly constantly now, with the occasional momentary drop to 50. I'll edit my original post as I find out which issues still apply.

WWarrengbrn 2018-08-26 github

My performance seems to be about 70% of what I get on windows 10. The frame rates are playable, but the frame times spike up very high every couple of seconds making navigating around the map a choppy experience. Every red tick on the graph was a moment where the game stuttered briefly.
ye

Ubuntu 18.04
Ryzen 1700
GTX 1070 with driver 396.54

AArekPiekarz 2018-08-26 github

@Warrengbrn What software are you using to display these stats? Or are they built into the game?

SSamStrongg 2018-08-26 github

Rotfiends are invisible because Stream Output is not supported by Vulkan.
https://github.com/doitsujin/dxvk/issues/135

Vvaleth 2018-08-26 github

@ArekPiekarz Those can be enabled for DXVK via an environment variable
https://github.com/doitsujin/dxvk#hud

WWarrengbrn 2018-08-27 github

@ArekPiekarz

/home/'user'/.steam/steam/steamapps/common/'Proton folder'

Rename user_settings.sample.py to user_settings.py and you can configure the performance hud from that file with the comments listed in the link @valeth sent.

?ghost 2018-08-31 github

Nvidia HairWorks does not work.

JJimmy1500 2018-09-04 github

Nvidia HairWorks worked for a week after updated Nvidia driver to 396.54 then stop working.

OS: Ubuntu 18.04 LTS
Graphic card: Geforce GTX 1070 (driver 396.54)
CPU: i7-8700 @ 3.70 Ghz

Aahallrq 2018-09-09 github

Is anyone else having no audio at all with TW3? I'm not sure if I've somehow mangled my audio or something but the game plays no sound whatsoever.

Fedora 28
Ryzen 7 1700 / GTX 1080 (driver 396.45) - probably not relevant but whatever
pulseaudio 12.1-1, alsa-lib 1.1.6-2

Ttriffid 2018-09-10 github

Audio works fine for me on both Mint 18.3 and Gentoo..

I am suddenly having a weird issue though, last night TW3 worked fine but today it deadlocks a few seconds after loading my game.. I have to force kill it to regain control of my desktop

Doesn't do that in Wine-3.15+dxvk-0.71, but the performance is much poorer for some reason.

Log attached, freeze happens at time ~49412 where it starts loading and unloading dinput a few times. Using Proton 3.7-6 Beta
steam-292030.log

TTuuvimer 2018-09-15 github

Audio will corrupt occasionally, restart of game 'fixes' it.
Invisible enemies are not just related to vulkan stream output not supported (https://github.com/doitsujin/dxvk/issues/135), there's also some oddities occuring when running with wined3d (some rotfiends are invisible, others are not, hounds of the hunt are invisible, in both dxvk and wined3d).

Odd keyboard inconsistencies when running with vulkan compared to PROTON_USE_WINED3D11=1 mode - I'd like someone to confirm or reject this:

In the game context, casting a spell is by pressing Q, once an 'alternative spell mode' is active and equipped, you hold Q to cast the alternative mode. When I run in wined3d mode, this works. When I run without it (dxvk/vulkan translation), for some reason this doesn't work. While holding the key, it just casts it normally. Tested so far is window vs fullscreen vs borderless etc, no change. (there are no ingame key bind options for changing this, unless I missed it)

Update: I ended up reproducing an old game bug that's still around.
Reference and fix: CD Project Red forum thread on it.

Adding a 'bug' here, I don't know what it is but it's consistent (on my system):
Running the game through the steam client will at some point reach a state that I thought initially was a memory leak (jittery/laggy fps that never clears). Config file settings for increasing memory/texture memory didn't make any difference. It always happens (no exceptions), during a long play session.

In an attempt to track down at least a nice error log, I then ran it with PROTON_DUMP_DEBUG_COMMANDS=1 %command%, and ran it via the run script it generates. The 'problem' was, it never occurs again. I then played about 20 (?) combined hours with it using the run script, never happens. Figuring a sneak update and bad timing on my part, I ran it in the steam client interface again, just now. 'Bug' hit about 20 minutes in.

There's an FPS drop (towards needing a restart to clear, not something you'd want to play with) due to unknown reasons that only occurs when running it through the steam client at random times (20 minutes in if you're lucky, looking to find it, or an hour or more in), that does not appear at all when running it in a terminal with the run script generated through the debug command option.

Something about what steam hooks into it, or the steam API <-> proton <-> this game? I'm clueless.
(Steam Beta client, have not tested this in non-beta)

SStructByLightning 2018-09-15 github

I get the staticky audio corruption bug too.

I'm not sure what my settings are but under the default plug and play I can use the normal mode and alt mode just fine.

Aahallrq 2018-09-15 github

@triffid @Tuuvimer Cheers for the responses. Seems as it may just be an issue with my setup as it now appears to happen in some other games. Maybe a bug in Fedora or something. That said, when I do get sound in TW3 it's not too staticy and I haven't gotten any major audio corruption yet.

?ghost 2018-09-16 github

The Earth Elemental (Contract: Doors Slamming Shut) is invisible.

Mmichaldybczak 2018-09-20 github

Audio will corrupt occasionally, restart of game 'fixes' it.

It happens to me sometimes too and I noticed, it's a depends on... Steam Controller being on. When I disable Steam Controller, restart steam, launch the game, sound issue is gone and then I can enable Steam Controller again. Weirdly enough, this happens only rarely. For the most time, Steam Controller or Big Picture Mode are fine and are not causing a sound issue. So all in all, this is an easy workaround.

My experience with Witcher 3 is:

  • I have in average 20% lower FPS than on Win10, which is expected by dxvk so it's all right
  • lags can happen sometimes but it's a matter of hardware catching up, so that's not a software related in my case, 99% of time game is fluent and 100% playable with ultra settings on gtx 970M (Manjaro), which is subjectively comparable to windows, so it's almost like native-like experience, when one forgets stats and some minor bugs
  • nvidia hairworks don't work, not a big deal in the end effect, it would be nice if it worked but it's not crucial
  • some monsters are invisible, also not a big deal, they can still be easily defeated, so for me, those are high-level earth elementals (low-level ones are visible) and sirens. Maybe once or twice, I had some other rare monster being invisible. Other than that it's all fine.
  • in some locations water animation glitches, so I can see straight cracks of the surface and beneath is as there was no water, but for most of the time water is rendered correctly
  • the game may freeze but that happened twice for a 2-week gameplay so that's like a non-issue

In my opinion, the game is perfectly playable on Proton. There are some glitches here and there that may be ironed out with time but all in all, it's a fantastic experience.

Ttriffid 2018-09-20 github

I use neither steam controller nor big picture mode, and my sound will become gritty and echoey after several hours of gameplay (exact time varies). Simply restarting TW3 fixes it.

The sound issue is thus not related to steam controller or steam big picture mode.

Mmichaldybczak 2018-09-20 github

@triffid, then we have different sound issues. Once the game launches with correct sound, it stays ok all the time.
It would be nice to see if some kind of bugs happen more often on certain distros then others. This could tell us if the issue is with package versions.
From what I saw, Ubuntu users on average have more issues with Witcher 3 than Arch/Manjaro ones but that may be a flawed observation.

?ghost 2018-09-29 github
  • Trees shake too much in Linux
  • Higher GPU fan noise in Linux compared to Windows (GPU: R9 390, FPS: locked at 30)
JJimmy1500 2018-10-12 github

Regarding the staticky audio corruption issue, here is what I found:

  1. Playing a game session continuously for extended period of time (more than 2-3 hours for me, but I imagine this can vary depending on the machine) consistently reproduces the issue.
  2. Same issue can be reliably reproduced using the method above in more than one games (Witcher 3, Endless Space 2)
  3. Like many people mentioned before, for both games in which I found the issue, it can be reliably "resolved" (temporarily at least) by restarting the game session one way or another, until the ongoing session lasted extended period of time again.

If I am to take a wild guess, it sure feels like some kind of memory buffer overflow problem to me.

Peripheries of my machine:
Ubuntu 18.04.1 LTS
GeForce GTX 1080 Driver Version: 396.54
Intel i7-8700

Mmichaldybczak 2018-10-13 github

The audio issue seems to be gone. Anyway, on Manjaro I haven't experienced the audio problem since I described it here (23 days ago). I was playing quite often recently (just finished the game) and audio was fine. There were so many Witcher 3, Steam and Proton updates so I assumed this problem was fixed during one of them.

Also, it looks like the problem was not identical to all people. For me, it was quite opposite to what was described above. The issue happened only at the beginning, but when I managed to run the game correctly, it was just fine no matter how long the game was on. But the sounds were "staticky" so the description is accurate.

So basically the only issue left for me was the lack of some monsters or even faces (rarely). The water issue was gone as well.

Once in a while game crashed after a load but as I recall games in overall were also crashy on Windows, so one crash for 2 days is not standing out ;) - everything within a norm IMO.

In general Witcher 3 runs great. Just fix the missing textures bug and all is fine.

Ddoitsujin maintainer 2018-10-13 github

@michaldybczak what "missing textures" bug? That description is rather ambiguous, people call literally everything "missing textures".

Invisible monsters should be fixed with Proton 3.16 if you have a Vulkan driver that supports the new Transform Feedback extension.

Mmichaldybczak 2018-10-13 github

Missing textures aka invisible monsters (sometimes vampire faces).

Thanks for the info about when the bug will be squashed. I use Manjaro so I will rather have a driver that will support it, providing the driver support for it will show at the same time.

Hhswhite33 2018-10-14 github

My $0.02 using Proton 3.16-1 beta

  • Mouse sensitivity is unplayably high - moving the mouse the smallest amount possible to the left or right spins the camera almost 360 degrees (with the in-game mouse sensitivity turned down to the minimum). For reference I'm using a Logitech G502 and I've tested at 125Hz and 1000Hz polling rates and a range of DPI settings
  • FPS is noticeably lower than in Windows, but still playable
  • Geralt's hair doesn't render

I would have tested some more (combat, etc) but the mouse issue makes it completely unplayable. Worth mentioning that this particular issue did not occur in Proton 3.7

System info

Ttomaszg7 2018-10-14 github

It won't start for me with Proton 3.16-1 beta and Nvidia 410.57. Worked quite well with older version.

HHanPrower 2018-10-14 github

Proton 3.16-1 beta works well for me and everything seems to be rendering everything correctly after patching mesa with the RADV patches for VK_EXT_transform_feedback.

Like others, however, I do get the sound going crackly sometimes. Sometimes after hours of play, sometimes only in cutscenes, sometimes only on the main menu. Relaunching the game sometimes fixes the issue, sometimes the issue just moves to another part of the game.

I haven't figured out any specific reason, or any events that cause it.

System Info

Bbarni2000 2018-10-14 github

@HanFox try this PULSE_LATENCY_MSEC=60 %command%

Mmichaldybczak 2018-10-14 github

Just checked Proton 3.16-1 beta, nvidia 410.57-2, using Steam Controller (so I can't say anything on mouse sensitivity that was posted above), OS: Manjaro KDE unstable, kernel 4.18.14-1.
Witcher 3 downloaded two large files after switching to Proton 3.16-1 beta so I expected some changes.

Everything seems to be correct but Sirens are still invisible, so maybe the driver is not there yet? So in the end nothing new. Sound is ok.

Ddoitsujin maintainer 2018-10-14 github

@michaldybczak You need the 396.54.09 driver for transform feedback support.

Mmichaldybczak 2018-10-14 github

I don't get it. I need an older driver to get it to work?

Ddoitsujin maintainer 2018-10-14 github

It's not an older driver, it's the newest Vulkan beta driver that came out yesterday.

Mmichaldybczak 2018-10-14 github

I thought that vulkan driver is within nvidia driver. If not, I'm not sure what package I should check on the Arch-based system. My packages are on Arch stable level.

EDIT: I checked with Arch wiki, vulkan driver for nvidia is within nvidia driver so I'm not sure where you get the number of the vulkan driver alone? Which nvidia driver will support that? If it came out yesterday then maybe it's still in arch unstable or testing so I should get it soon.

Kkisak-valve maintainer 2018-10-14 github

Hello @michaldybczak, nVidia has a vulkan developer series for experimental features at https://developer.nvidia.com/vulkan-driver. This driver series doesn't need to pass their internal quality control and the purpose is to get early feedback on new things before they get integrated into the mainstream driver.

Mmichaldybczak 2018-10-14 github

So that means I just have to wait for it a little longer. I won't mess with GPU drivers by using developer driver, so I just update for the version that will come with this feature when it shows.
Thanks for the info @kisak-valve.

HHanPrower 2018-10-14 github

@HanFox try this PULSE_LATENCY_MSEC=60 %command%

So far so good. Thanks @barni2000

Ttomaszg7 2018-10-14 github

@HanFox try this PULSE_LATENCY_MSEC=60 %command%

I also suffer from audio distortion in cinematics, but in my case this doesn't help. I don't have Pulse audio installed on my system (tried on Proton 3.7-8, as newer one is broken for me).

System info

Bbit-stone 2018-10-15 github

Hey there,
don't know if I am right here.
I was able to play the game without problems until recently.
Only had the audio distortion problem after a couple of hours of playtime.

But now I cant even start the game.
First I can see some text appearing (Witcher 3 is a trademark ...) but with a very low framerate.
After that I only see the "Skip" in the lower right corner.
But nothing happens. I need to kill the process in order to get out.

Any ideas?
System

Ttomaszg7 2018-10-18 github

It won't start for me with Proton 3.16-1 beta and Nvidia 410.57. Worked quite well with older version.

FWIW, recent updates to Proton 3.16-2 and -3 fixed the problem for me.

Xxexpanderx 2018-10-19 github

I could launch the game before, now it wont start (but it says it is running in Steam).

This is my spec:
GPU: GTX 1080
Nvidia driver: 410.66
Kernel version: 4.14.77
Proton version: 3.7 or 3.16-3

However, I can run the game via Lutris using wine-3.18-staging and DXVK 0.90.

How can I run proton via the command line to see if I can see any errors?

Mmichaldybczak 2018-10-19 github

There is a way to enable logs for certain Steam Play games. I don't remember exactly how it is done (I did it and now I can't undo it ;p) but maybe someone hare knowns this. I only remember I had to find some proton config for the given prefix and edit something there.
Alternatively, just run Steam in a terminal and then launch WT3.

For me, the game starts without a problem, although the first launch after an update is always visibly longer. I use GTX 970M and kernel 4.18, rest is the same.

Ppchome 2018-10-19 github

@michaldybczak

I don't remember exactly how it is done ...

You shouldn't memorize everything, but one thing: use README (man, info, wiki, google ...).
https://github.com/ValveSoftware/Proton#runtime-config-options

Xxexpanderx 2018-10-19 github

Well, it is not working with Proton. I attach the log if someone can see something in there....
steam-292030.log

It works well with wine-staging and dxvk, not with Proton.

My xorg-server version is: 1.20.2 (if it has something to do with this).

Kkisak-valve maintainer 2018-10-19 github

Hello @xexpanderx, please copy your system information from steam (Steam -> Help -> System Information) and put it in a gist, then include a link to the gist in this issue report. Additionally, does ulimit -Hn output a high value or 4096?

Ppchome 2018-10-19 github

@xexpanderx
It is working with Proton, and past several days I'm enjoying Blood&Wine New game on Hardcore difficulty.
No special settings required, game starts flawlessly from first ever run using Proton.

Your logs contains strange entries:

Fontconfig warning: "/etc/fonts/conf.d/60-latin.conf", line 4: unknown element "its:rules"
Fontconfig warning: "/etc/fonts/conf.d/60-latin.conf", line 5: unknown element "its:translateRule"
Fontconfig warning: "/etc/fonts/conf.d/60-latin.conf", line 8: unknown element "description"
18997.120:0024:0025:warn:module:load_builtin_dll failed to load .so lib for builtin L"Z:\\home\\alexsson\\.local\\share\\Steam\\steamapps\\common\\The Witcher 3\\bin\\x64\\D3DCOMPILER_47.dll": /home/alexsson/.local/share/Steam/steamapps/compatdata/292030/pfx/dosdevices/z:/home/alexsson/.local/share/Steam/steamapps/common/The Witcher 3/bin/x64/d3dcompiler_47.dll: invalid ELF header
18997.123:0024:0025:warn:module:load_builtin_dll failed to load .so lib for builtin L"Z:\\home\\alexsson\\.local\\share\\Steam\\steamapps\\common\\The Witcher 3\\bin\\x64\\PhysX3Common_x64.dll": /home/alexsson/.local/share/Steam/steamapps/compatdata/292030/pfx/dosdevices/z:/home/alexsson/.local/share/Steam/steamapps/common/The Witcher 3/bin/x64/PhysX3Common_x64.dll: invalid ELF header

Looks like your setup is somehow broken.

Xxexpanderx 2018-10-19 github

Hello @xexpanderx, please copy your system information from steam (Steam -> Help -> System Information) and put it in a gist, then include a link to the gist in this issue report. Additionally, does ulimit -Hn output a high value or 4096?

Hi, the link: https://gist.github.com/xexpanderx/c7dccaf2d088729fa904e499f1212cc1

ulimit -Hn
1048576
Ppchome 2018-10-20 github

BTW, @kisak-valve
do you really require WINEDEBUG=+module ?

this produces either more noise than help
WINEDEBUG=+loaddll should produce more compact and user-frendly output.

Fflibitijibibo 2018-10-23 github

@kisak-valve Requesting an XAudio2 label for this game... this is a Wwise title but they still use an XAudio2 context for the output stream.

ZZoetrap 2018-10-31 github

Just installed The Witcher 3 and played about an hour in White Orchard. Everything worked just fine out of the box without any tweaking whatsoever. Ran quite well too at 1080p medium settings.

My specs:
https://gist.github.com/Zoetrap/3f0fc6d113c5fef9cdf7d623ea5e81c8

Sound card says "Realtek" in the Steam system info (which is the built-in soundcard), but I'm actually using an external USB Interface (Steinberg UR242) for all audio. Worked flawlessly during my testing.

MMikaelBox 2018-11-01 github

Proton 3.16-4, the game isn't start. 3.7.8 - all works

Kkisak-valve maintainer 2018-11-01 github

Hello @MikaelBox, please copy your system information from steam (Steam -> Help -> System Information) and put it in a gist, then include a link to the gist in this issue report. Additionally, please add PROTON_LOG=1 %command% to the game's launch options and drag and drop the generated $HOME/steam-$APPID.log into the comment box.

MMikaelBox 2018-11-01 github

Hello @MikaelBox, please copy your system information from steam (Steam -> Help -> System Information) and put it in a gist, then include a link to the gist in this issue report. Additionally, please add PROTON_LOG=1 %command% to the game's launch options and drag and drop the generated $HOME/steam-$APPID.log into the comment box.

https://gist.github.com/MikaelBox/5d3acc771f4f92e3cde8a3fbf7204e70
I tried to add to launch options PROTON_LOG=1 %command%, but steam-292030.log is empty

Mmichaldybczak 2018-11-02 github

I enabled logging by changing some proton config file directly. I don't remember exactly but I had to find proton's folder for Witcher3 and then choose some specified file there and change some line in it... I'm beating myself I didn't write any note :(.

Xxexpanderx 2018-11-02 github

I found the solution for my problems. I fixed it by this:

export STEAM_RUNTIME_PREFER_HOST_LIBRARIES=1

This setting prefer my system libraries.

Ggerardlouw 2018-11-04 github

Tested with Proton 3.16-4 Beta on Kubuntu 18.04.

When using the mouse to control the camera, I am unable to rotate the camera past a certain point when the (invisible) mouse hits any edge of the screen. Problem persists through Full Screen, Windowed, and Borderless Windowed modes.

Bbit-stone 2018-11-04 github

@gerardlouw
I had a similar issue - fixed it by alt-tabbing a few times. Or by using windows-key and clicking on it (used gnome)

I miss playing the game. It ran just fine and now it wont start :/

Ttriffid 2018-11-05 github

When using the mouse to control the camera, I am unable to rotate the camera past a certain point when the (invisible) mouse hits any edge of the screen. Problem persists through Full Screen, Windowed, and Borderless Windowed modes.

This happened to me a lot more often in previous versions of proton than the current one.

The fix I found is press esc to bring up the game menu, go into options, wave your mouse so the cursor hits both sides of the screen, then return to the game and it should be fixed.

Simply pressing esc to open the game menu isn't sufficient, have to actually click into something. Not sure if waving the mouse is necessary, that may be superstition.

Ggerardlouw 2018-11-05 github

@kzoltner @triffid
Thanks, good to know it's not just me. It got too frustrating so my workaround was to play with a controller instead.

Ttriffid 2018-11-05 github

With latest proton, sound still develops echoes and crackle after a few hours of gameplay.

Really sounds like a buffer mismanagement issue, as if witcher has its own timer for buffer updates, ignoring information from the soundcard and, after a while, it ends up swapping buffers at the wrong time or something.

I'm not sure why this causes the same sound sample to be played a second or third time with maybe a couple hundred ms lag overlaid with the first, but that's what it sounds like is happening.

Kkisak-valve maintainer 2018-11-06 github

Witcher 3 - can't use kb/mouse

Issue transferred from https://github.com/ValveSoftware/Proton/issues/1927.
@qsecofr posted on 2018-11-06T18:50:09:

Compatibility Report

  • Name of the game with compatibility issues: The Witcher 3: Wild Hunt
  • Steam AppID of the game: 292030

System Information

I confirm:

  • [ ] that I haven't found an existing compatibility report for this game.
  • [x] that I have checked whether there are updates for my system available.

steam-292030.log

Symptoms

When I launch Witcher 3 I am stuck with the controller control scheme, Keyboard assignements are empty and I can't use kb/mouse to control the character nor interact with the game. Big Picture shows controller icon.
I opened a ticket with steam support (Ticket: HT-VQJ4-RCXT-865Y), they pointed me to github/proton.

Reproduction

Everytime I start the game

Kkisak-valve maintainer 2018-11-06 github

Witcher 3 (292030) Vega 64 support

Issue transferred from https://github.com/ValveSoftware/Proton/issues/1929.
@bogdans41472 posted on 2018-11-06T22:07:45:

Compatibility Report

  • Witcher 3 Wild Hunt
  • Steam AppID of the game: 292030

System Information

I confirm:

  • [x] that I haven't found an existing compatibility report for this game.
  • [x] that I have checked whether there are updates for my system available.

Symptoms

When loading save games, Wine stops responding.

Reproduction

  1. Start game
  2. Continue story
  3. Wait for story cinematic
  4. Loading bar is displayed
  5. Observe freeze

Note, I have also tried installing mesa-vulkan-drivers and vulkan-utils from the default ubuntu ppa, but that didn't help at all, so I stuck with the latest drivers from AMD - 18.4.

Additionally, I have also tried starting a new game, but the same issue occurred after the beginning cinematic.

Kkisak-valve maintainer 2018-11-06 github

Hello @bogdans41472, you may have better luck with a build of mesa that supports VK_EXT_transform_feedback and built against llvm 7 or newer (like padoka unstable, oibaf, steamvr experimental PPAs)

GGloriousEggroll 2018-11-07 github

@bogdans41472 Witcher 3 on Vega 64 needs latest llvm 7 and mesa 18.3. DXVK (part of proton) officially requires at least mesa 18.2 and llvm 7, and does not work with amdgpu-pro(amd official) drivers, as the vulkan portion is several months outdated: https://github.com/doitsujin/dxvk/wiki/Driver-support . The reason 18.3 is required is because Witcher 3 needs transform feedback, which was not included in 18.2

source: I own a vega 64 and have been active in the DXVK community for several months.

Also your gist log shows page not found.

Bbogdans41472 2018-11-07 github

Hi @kisak-valve and @GloriousEggroll - I uninstalled the amdgpu-pro drivers and installed the ones provided in this ppa: ppa:paulo-miguel-dias/pkppa . Everything is running smoothly now :+1:. Thanks for your help.

By the way, I really want to help out with proton development/testing. What would you guys recommend? What's the best way I can help? I'm currently working as a dev in test for a security company, but I have time in the evenings/weekends to test things around.

Kkisak-valve maintainer 2018-11-07 github

In general, the best way to help Proton is improve upstream wine and improvements there will trickle downstream.

Kkisak-valve maintainer 2018-11-12 github

Witcher 3 not running when esync is used

Issue transferred from https://github.com/ValveSoftware/Proton/issues/1954.
@kzoltner posted on 2018-11-11T23:12:34:

System Information

PROTON_LOG

https://gist.github.com/kzoltner/280a6af0acf5b561e924171aa86fa8cc

Symptoms

Hey there,

I do have a small problem with Witcher 3 on Steam Play.
A little while ago it worked out of the box without me changing anything. Had similar performance to Windows.

Today I wanted to play again but I was unable to get it working.

I start the game, wait a few seconds. Screen turns black and I get quite some load on my CPU.
I can see some copyright text appear at about 0.5 frames per second. I can skip that but the game crashes.

If I disable ESYNC by using
PROTON_NO_ESYNC=1 %command%
it does start without any problems but I only get around 15 to 35 FPS and sometimes during a fight the game just locks up and I need to restart my computer.

Kkisak-valve maintainer 2018-11-18 github

Witcher 3 (292030) - black screen :/

Issue transferred from https://github.com/ValveSoftware/Proton/issues/1997.
@rafik777 posted on 2018-11-18T17:41:18:

Compatibility Report

  • Name of the game with compatibility issues: Witcher 3
  • Steam AppID of the game: 292030

System Information

  • GPU: GTX 850M
  • Driver/LLVM version: nvidia 410.73
  • Kernel version: 4.15.0-39-generic
  • Link to full system information report: Gist system information
  • Proton version: 3.16

Symptoms

After I start the game I see only black screen (no sound too).
Only thing I can do is killing the process.

steam-292030.log

Llieff 2018-11-18 github

@rafik777 Try 396.54.09 driver, 410.73 have less Vulkan features. At least

err:   D3D11: CreateGeometryShaderWithStreamOutput:
Transform feedback not supoorted by device

should go away.

Ppchome 2018-11-18 github

@rafik777

AL lib: (WW) GetSymbol: Failed to load jack_error_callback: /usr/lib/x86_64-linux-gnu/libjack.so.0: undefined symbol: jack_error_callback
AL lib: (WW) jack_msg_handler: Cannot connect to server socket err = No such file or directory
AL lib: (WW) jack_msg_handler: Cannot connect to server request channel
AL lib: (WW) jack_msg_handler: jack server is not running or cannot be started
AL lib: (WW) jack_msg_handler: JackShmReadWritePtr::~JackShmReadWritePtr - Init not done for -1, skipping unlock
AL lib: (WW) jack_msg_handler: JackShmReadWritePtr::~JackShmReadWritePtr - Init not done for -1, skipping unlock
AL lib: (WW) ALCjackBackendFactory_init: jack_client_open() failed, 0x11
AL lib: (WW) alc_initconfig: Failed to initialize backend "jack"

Check your ~/.alsoftrc, and set e.g.

[general]
drivers=pulse,alsa
;drivers=pulse,alsa,jack

Alternatively you can try to disable xaudio2_7.dll.
Or LD_PRELOAD=/path/to/correct/libjack.so.0.

Rrafik777 2018-11-18 github

@pchome Here https://github.com/ValveSoftware/Proton/issues/607#issuecomment-419906100
I see the same problem with libjack in logs and Witcher 3 is playable.
I was struggling with this two month earlier and was moment that this ERROR was gone, but black screen stayed... :/

Rrafik777 2018-11-18 github

@pchome @lieff
Nvidia 396.54, Proton 3.7:
https://gist.github.com/rafik777/1a41055c3de932d894223102857d8a51

No jack errors, no D3D11 errors.
Still black screen... :/
Any ideas?

Llieff 2018-11-19 github

No idea, except also trying Proton 3.16-4 Beta (is it's not typo about 3.7) or wine 3.19/master (3.20 is known broken).

Aaldonogueira 2018-11-20 github

That dwarf banker, from Vivaldi bank in Novigrad, has invisible arms. I guess it's the only bug I still can see in this game. Tell me if you need a screenshot

Ubuntu 18.04.1, AMD Radeon RX 470, Mesa drivers from padoka PPA, vulkaninfo says it's Vulkan 1.1.70, Proton 3.16-4 Beta

Ttriffid 2018-11-21 github

He's had arms every time I've visited him, can check again later tonight in case there's a regression.

GTX1050Ti, nvidia-396.54.09, proton 3.16-something (have to check)

Llieff 2018-11-21 github

@aldonogueira For stream output you need Mesa 18.3-git. Ubuntu 18.04 do not have one out of box.
See https://github.com/doitsujin/dxvk/wiki/Driver-support

Ddoitsujin maintainer 2018-11-21 github

@aldonogueira which Mesa and LLVM version is that exactly? With both latest mesa-git and llvm-svn, the arms render fine on my end.

Aaldonogueira 2018-11-21 github

Stream output is working. I can see the fancy new hair on Geralt, the rotfiends, the sirens and the wild hunt dogs. I guess it's another problem.

According to glxinfo, it's using Mesa 19.0.0-devel - padoka PPA, LLVM 8.0.0. And vulkaninfo informs that VK_EXT_transform_feedback is available. It says the version is 1.1.70, but apparently there are also some extensions from 1.1.73.

@triffid @lieff @doitsujin

Mmichaldybczak 2018-11-21 github

Thanks for the info, I still have on Manjaro testing:
Mesa 18.2.5-0
nvidia 1:410.73-1
steam beta
proton 3.16-4
kernel 4.19
GPU: nvidia gtx 970M + intel, game started on nvidia-xrun for vulkan compatibility

Witcher 3 runs fine as always but the above-mentioned monsters are still invisible. I guess I have to wait till Mesa 19 (or 18.3?) and higher nvidia driver.
Dwarf banker in Novigrad always had arms so that has to be something else.
Aside few invisible monsters (which isn't that problematic for gameplay), some vampires in Blood and Wine, especially their heads or faces at end of the main quest are missing on cutscenes (not always). I believe one character also had an invisible face in one of the cutscenes. So the bug is showing here and there but already said, it's not spoiling the fun. However, it would be great to see that issue to be gone. When that happens, then I really don't see much issue with TW3, all works like on Windows (minus 15% FPS), so I'm very happy because of that. I already finished the game and spend many hours in it on Linux to know how well it works.
Ah, forgot nvidia hairworks. This is not important but out of curiosity, does it work now or will the support also come with the strep output?

Ddlove67 2018-11-21 github

@michaldybczak Nvidia doesn't use the Mesa driver (though your intel chip does), you can update to 396.54.09 here: https://developer.nvidia.com/vulkan-driver which support transform feedback.

Once you do that, all of those monsters should be visible, and you should be able to use hairworks.

Mmati865 2018-11-21 github

@dlove67 that link won't work for 4.19 kernel but there are patched drivers for Arch based distros: https://aur.archlinux.org/packages/nvidia-vulkan-dkms

Mmichaldybczak 2018-11-21 github

@dlove67, I'm on 410 version and it's not easy to downgrade to older series without messing the system. Manjaro is a rolling distro and has its own, unique packages for nvidia drivers (suited to mhwd scrips), so I would have to compile them myself and know how to install them without breaking my desktop. So even arch drivers aren't suited (directly) for Manjaro.

I rather wait till that support will show up in 410 or newer series. Sooner or later it will come.

Ddoitsujin maintainer 2018-11-21 github

@michaldybczak There is an AUR package somewhere for 396.54.09 (nvidia-vulkan). As of right now there is no ETA on when Nvidia will merge transform feedback support into their mainline drivers.

@aldonogueira That's weird. Can you post a screenshot?

Mmichaldybczak 2018-11-21 github

@doitsujin, installing this package would break Manjaro for sure. As already said, Manjaro uses different packages for gpu drivers than Arch. Yes, drivers are taken from Arch but renamed and repackaged to suit Manjaro specific hardware detection/installation/management tool - MHWD which currently supports only two nvidia driver lines: the newest one (at moment of writing this is 410xx) and 390xx (for support of some fenri based cards).

Aaldonogueira 2018-11-21 github

Here is the armless dwarf @doitsujin
image

In this screenshot the game effects was set to medium

Xxexpanderx 2018-11-21 github

I can confirm I have the same problem, an armless dwarf.

Nnottux 2018-11-25 github

When i launch the game , game process freezes (%0 cpu usage). but once when i left it open for steam cards and gone to dinner, game was launched somehow but when i try to restart after setting graphics settings, it didn't restated
Games borked on me:
The Witcher Enhanced Edition (Lutris version works but can't launch The Witcher camping, stucks at loading, can launch the sidemissons)
The Witcher 3 (Lutris version do work)
Warframe: applied Glorious Eggrolls patch but stuck at loading screen, it used to work issue page

System Information

Lutris version works with esync but game was choppy there until i disabled sharpening effect from post processing effects

Sshmerl 2018-12-07 github

@aldonogueira : can you post a save with the armless dwarf in this thread please? I'd be interested in testing it using radv and upstream Wine+dxvk.

NNoXPhasma 2018-12-12 github

With Proton 3.16-5 it's using FAudio and it works. However, it only uses Stereo (The same is for Prey (2017) btw), while the game is capable of surround and so is my system configuration. Is there a way to tell FAudio to actually respect my Pulseaudio configuration and use it? @flibitijibibo

I've already tried to set format = AL_FORMAT_51CHN16 in the ~/.alsoftrc, but this has no effect.

Fflibitijibibo 2018-12-12 github

SDL_AUDIO_CHANNELS=6 %command% should fix that.

NNoXPhasma 2018-12-12 github

It does indeed, thanks.

Mmichaldybczak 2018-12-18 github

I just tested Witcher 3 on nvidia 1:415.22-1 with added Transform Feedback on Proton 3.16-5-Beta and invisible monsters (like Sirens) are finally visible! Now Witcher 3 is 100% working! Awesome!

Ttastyminerals 2019-01-03 github

@michaldybczak can you please provide a link or a solution how to add "Transform Feedback"?

NNoXPhasma 2019-01-03 github

@tastyminerals Transform Feedback is a DirectX function, which was added into Vulkan 1.1.88 and DXVK 0.81. What you need is a driver for your graphics card which supports Vulkan 1.1.88 and Proton Version >= 3.16-1.

Mmichaldybczak 2019-01-03 github

@michaldybczak can you please provide a link or a solution how to add "Transform Feedback"?

As @NoXPhasma said, you only need to have appropriate gpu driver version and newest Proton beta and it should work out of the box.

Sshmerl 2019-01-03 github

Too add some details, in case of Mesa you need 18.3.0 and newer to have VK_EXT_transform_feedback.

Nnslynx 2019-01-04 github

Have a problem: witcher 3 cannot save settings and each time it said, that it is starting for the first time
Anyone can reproduce it?

Sshmerl 2019-01-04 github

It works fine in upstream Wine+dxvk (GOG GOTY version). So something must be specific to your setup.

Nnslynx 2019-01-04 github

Witcher save progress in bin (??) directory, but cannot find settings there. So I can save/load game, but have to set settings after every launch.
Path ~/.steam/steam/steamapps/compatdata//pfx/drive_c/users/steamuser is empty

Every time I got file in Temp directory like this:
[003C:003D][2019-01-05T00:26:19]i001: Burn v3.7.2829.0, Windows v6.1 (Build 7601: Service Pack 1), path: Z:\home\kolya.steam\steam\steamapps\common\The Witcher 3_CommonRedist\vcredist\2013\vcredist_x64.exe, cmdline: '/quiet /norestart'
[003C:003D][2019-01-05T00:26:19]e000: Error 0x80070002: Failed to find local per-machine appdata directory.
[003C:003D][2019-01-05T00:26:19]e000: Error 0x80070002: Failed to build cache directory.
[003C:003D][2019-01-05T00:26:19]e000: Error 0x80070002: Failed to set registration paths.
[003C:003D][2019-01-05T00:26:19]e000: Error 0x80070002: Failed to parse registration.
[003C:003D][2019-01-05T00:26:19]e000: Error 0x80070002: Failed to load manifest.
[003C:003D][2019-01-05T00:26:19]e000: Error 0x80070002: Failed to initialize core.

Mmichaldybczak 2019-01-04 github

Check correctness of the installation. If it claims to be OK, delete something like (proton prefix for Witcher 3) and let Steam download missing files. See if it helped.

Nnslynx 2019-01-05 github

I have switched my distro from debian-stable based to ubuntu 18.04 (without cleaning /home directory, of cause), may be I should clean some settings?

Ggrigi 2019-01-05 github

I have been playing The Witcher 3 on intel ANV (Kabylake Core i7-8650U on a Dell notebook with a UHD620).
I'm running Mesa git (19) + VK_EXT_transform_feedback patches on Gentoo 64-bit.
Using unpatched proton 3.16-6 Beta from within Steam. KDE5 desktop.

The game actually runs very playable (on lowest settings, but medium texture resolution @ 1280x720), but it has the following issues:

  • Some textures occasionally go black, or rather much darker. This tends to get worse the longer one plays but not always. It makes navigating caves very difficult.
  • Sometimes a texture will flicker between correct and dark.
  • Generally too dark,
  • Occasionally lips don't sync with the voices (I suspect it happens when framerate drops too low)

The good news I'm seeing rotfiends, wild hunt hounds, etc... and have no funky polygon distortion or missing polygons, only some texture issues.

Ggrigi 2019-01-07 github

I did some more experimentation, the darkening appears much like a shadow.
Easy to spot on a wall of a building outside that is partly in shadow, it will flicker between having a shadow, and looking like it is all in shadow.
Also easy to spot on rocks, where the whole rock will appear in shadow just because a part is shaded.
Do you need more info? e.g. screenshots?
Should I test if this issue is happening on mesa 18.3 as well? (instead of just git)

Ggrigi 2019-01-08 github

Here is a video recording of the issue if you need some clarity:
https://drive.google.com/open?id=1MohQ4EGNQ94PDaS6a9wbjy5yIdYirE5G

Kkisak-valve maintainer 2019-01-24 github

Witcher 3 performance regression with 3.16-6

Issue transferred from https://github.com/ValveSoftware/Proton/issues/2262.
@sirkon posted on 2019-01-24T08:27:26:

There's noticeably performance regression in Witcher 3 at least in Toussaint: looking around with mouse doesn't smooth anymore although it was with before Proton 3.16 (didn't try with older Protons 3.16-X)

System info: https://gist.github.com/sirkon/601db826009f1fb02ac48ac0c5e87c6e

?ghost 2019-01-27 github

FPS is sporadic with 3.16.6 beta. Seems to jump from 60 with drops into 40's and even single digits at times. It is not prolonged and the game is playable, just sporadic and choppy with FPS.

System info | https://gist.github.com/Makersmarx/25c73636cb80bb8b0b770dec85da7fe6

Sshmerl 2019-01-27 github

Just for the reference, with regular latest Wine+ latest dxvk the game is running pretty smoothly.

Ppclouds 2019-02-07 github

Input (mouse & keyboard) is not working for me. Both mouse and keyboard work in the menu, but once entered the game, none works. Steam log is here https://gist.github.com/pclouds/ef41512b5293eb896518ad53f7f98897

The input.settings file in "My Documents/The Witcher 3" shows this. Version -1 can't be good but I don't know how to fix it. Perhaps I could get a good copy from someone to try out?

[InputSettings]
Version=-1

The key binding menu also shows none (except quick save) is bound. Everything is bound to "none". F5 (quick save) does indeed work. At least the game tells me I can't save at that point.

Ttriffid 2019-02-07 github

What happens if you delete the file and get the game to recreate it?

Mine's here if that doesn't work

Ppclouds 2019-02-08 github

The game always recreates the file with the same content. If I put your file in, it overwrites it too. If I remove write permission, then your file remains, but it does not reflect in the key binding menu (everything is still bound to None there) and no key or mouse works.

AAntiSC2 2019-02-10 github

I have the same problem as @tribunal2000, each time I launch it the settings resets. It actually writes the to the config file but it doesn't seem to read from there. Save games works and the only strange thing I notice is that it seems to try to update the game every launch.

System info:
Solus Linux Operating System
OpenGL vendor string: X.Org
OpenGL renderer string: Radeon RX Vega (VEGA10, DRM 3.27.0, 4.20.3-108.current, LLVM 7.0.0)
OpenGL core profile version string: 4.5 (Core Profile) Mesa 18.3.2
OpenGL core profile shading language version string: 4.50
OpenGL core profile context flags: (none)
OpenGL core profile profile mask: core profile
OpenGL core profile extensions:
OpenGL version string: 4.5 (Compatibility Profile) Mesa 18.3.2
OpenGL shading language version string: 4.50
OpenGL context flags: (none)
OpenGL profile mask: compatibility profile
OpenGL extensions:
OpenGL ES profile version string: OpenGL ES 3.2 Mesa 18.3.2
OpenGL ES profile shading language version string: OpenGL ES GLSL ES 3.20
OpenGL ES profile extensions:

uname -a:
Linux linuxbox 4.20.3-108.current #1 SMP PREEMPT Tue Jan 22 00:34:33 UTC 2019 x86_64 GNU/Linux

Proton 3.16-6 Beta

XXakepSDK 2019-02-11 github

On multihead setup, mouse will stop working at side of next monitor border.
E.g. i have main monitor and second, second is left of main, game runs on main, i move mouse left in game and after some time camera stops from rotating.

NNoXPhasma 2019-02-11 github

@XakepSDK I have the same monitor setup (2x 1080p, right one main) but not that issue. Neither with Borderless Window nor with Fullscreen.

XXakepSDK 2019-02-11 github

@NoXPhasma what DE do you have? Xorg or Xwayland?

NNoXPhasma 2019-02-11 github

Xorg
Cinnamon
Nvidia 415.27

Full Systemfino: https://gist.github.com/NoXPhasma/3c27464475ddcb3157a0cf4c53b707be

NNoXPhasma 2019-02-11 github

I have no experience with KDE, but maybe it's some setting in Kwin which prevents the mouse grab? Do you have that in other games running with proton?

XXakepSDK 2019-02-11 github

Yeah, disabling KWin seems helped me! Thanks.

Ttriffid 2019-02-11 github

Hmm I'm running KDE and no issues here.. only one screen though

On Mon, 11 Feb 2019 at 16:39, XakepSDK [email protected] wrote:

Yeah, disabling KWin seems helped me! Thanks.

—
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub
https://github.com/ValveSoftware/Proton/issues/607#issuecomment-462251133,
or mute the thread
https://github.com/notifications/unsubscribe-auth/AAKBGhC3kC9JM0jcDhgOTH7khoFemhBLks5vMSw0gaJpZM4WMl0z
.

NNoXPhasma 2019-02-11 github

It's obvious that you won't run into this issue, when you only have one monitor. As your mouse can't leave the game to the next monitor ;)

XXakepSDK 2019-02-11 github

Actually, my mouse didn't leave to another monitor, it's just stuck at leftmost position of main monitor.

Ttomaszg7 2019-02-11 github

I have a single monitor setup (with xfwm4) and from time to time the same issue with mouse happens as well. Usually alt-tabbing out and back into the game fixes it.

?ghost 2019-03-16 · hidden on GitHub github

Doesn't start with Proton 3.16-4 (no custom launch options used).

Kkisak-valve maintainer 2019-03-16 · hidden on GitHub github

Hello @atomsymbol, please opt into Steam's beta and/or go to Steam -> Library dropdown -> Tools and install Proton 3.16 from the list.

If that does not have an effect, please copy your system information from Steam (Steam -> Help -> System Information) and put it in a gist, then include a link to the gist in this issue report. Also, please add PROTON_LOG=1 %command% to the game's launch options and drag and drop the generated $HOME/steam-$APPID.log into the comment box.

?ghost 2019-03-16 · hidden on GitHub github

Hello @kisak-valve:

  1. Steam -> Library dropdown -> Tools and install Proton 3.16 from the list: no effect (initially although Proton 3.16 is installed it shows as not installed in the Tools list)
  2. Opt into Steam beta: no effect, empty steam-292030.log
  3. Delete "Proton 3.16" directory and redownload Proton 3.16: no effect, empty steam-292030.log
  4. Steam -> Library dropdown -> Tools and install Proton 3.16 Beta from the list, select Proton 3.16-8 Beta for all titles, restart Steam: game starts correctly
  5. Select Proton 3.16-4 (non-beta) for all titles, restart Steam: game starts correctly

Final result: Game now starts with both 3.16-4 and 3.16-8 Beta.

Thanks.

Kkisak-valve maintainer 2019-05-07 github

Hello @grigi, looks like you were using an experimental transform feedback patch set. Are you seeing the same rendering quirks with mesa 19.0+ now that ANV transform feedback support was mainlined?

NNewbytee 2019-06-16 github

The dripping effect in the cave beneath Bywaters is broken and is only present on parts on the screen, varying depending on how far up or down you look.

20190616105525_1
20190616110047_1
20190616110039_1

Proton 4.2-7, nvidia 418.74 with a GTX 970 on Fedora 30.

Ddoitsujin maintainer 2019-06-16 github

@Newbie13XD can you post a savegame of that location?

NNewbytee 2019-06-17 github

@doitsujin Glad you ask! Here you go: ManualSave_86d7e_7e32bc00_2b78738.zip

Kkisak-valve maintainer 2019-06-17 github

Witcher 3 on ubuntu with Steam play won't start

Issue transferred from https://github.com/ValveSoftware/Proton/issues/2817.
@milka123 posted on 2019-06-17T15:52:24:

Compatibility Report

  • Name of the game with compatibility issues: Witcher 3: Wild hunt
  • Steam AppID of the game: 292030

System Information

I confirm:

  • [x] that I haven't found an existing compatibility report for this game.
  • [x] that I have checked whether there are updates for my system available.

Symptoms

When I start the game, Steam client says it's running but nothing shows up. When I check system monitor I can see witcher3.exe in processes as running but nothing happens.
I couldn't find any more details about potential crashes in steam app log file.

Reproduction

Downloaded Witcher 3: Wild hunt through steam client and tried to run it

steam-292030.log

Sshmerl 2019-06-17 github

@Newbie13XD: I can see the issue on AMD too (Vega 56). But could it be a game bug? May be ask some Windows users to test it.

You can do it here.

NNewbytee 2019-06-17 github

@Newbie13XD: I can see the issue on AMD too (Vega 56). But could it be a game bug? May be ask some Windows users to test it.

You can do it here.

Could be. I don't have Windows installed on a PC decent enough to run Witcher 3 unfortunately, and right now I neither have the time to create a post there, so if you do it that would be greatly appreciated.

Might be able to ask a friend though, or perhaps doitsujin could do it. We'll see.

Sshmerl 2019-06-17 github

right now I neither have the time to create a post there, so if you do it that would be greatly appreciated.

Just for the reference, you can use your GOG credentials on the Witcher forum, in order not to create new users.

NNewbytee 2019-06-22 github

right now I neither have the time to create a post there, so if you do it that would be greatly appreciated.

Just for the reference, you can use your GOG credentials on the Witcher forum, in order not to create new users.

Did you (or anyone else that might see this, for that matter) create a post there? If not I'll get one going as soon as possible.

Sshmerl 2019-06-23 github

Did you (or anyone else that might see this, for that matter) create a post there?

I didn't yet. Please do, when you have a chance.

NNewbytee 2019-06-23 github

Did you (or anyone else that might see this, for that matter) create a post there?

I didn't yet. Please do, when you have a chance.

Created a thread here, seems it is a game bug (that also occurs on Windows).

PPeterMX 2019-06-23 github

Witcher 3 on ubuntu with Steam play won't start

Issue transferred from #2817.
@milka123 posted on 2019-06-17T15:52:24:

Compatibility Report

  • Name of the game with compatibility issues: Witcher 3: Wild hunt
  • Steam AppID of the game: 292030

System Information

I confirm:

  • [x] that I haven't found an existing compatibility report for this game.
  • [x] that I have checked whether there are updates for my system available.

Symptoms

When I start the game, Steam client says it's running but nothing shows up. When I check system monitor I can see witcher3.exe in processes as running but nothing happens.
I couldn't find any more details about potential crashes in steam app log file.

Reproduction

Downloaded Witcher 3: Wild hunt through steam client and tried to run it

steam-292030.log

I have same problem, but when I installed the game(a week ago) it ran OK in first day. I thought it stoped working because I pasted some nvapi dlls in proton folder to make GTA 4 works, but when I switched to an older proton(one without nvapi dlls) the Witcher 3 still didn't launch. I also switched from 24GB of ram to 16GB(just in case this matter)

Pptersilie 2019-06-26 github

I've been having audio issues (popping, cracking sound and low framerate) in the intro cut-scenes as well as the recap cut-scenes that play when loading a save-file. The issue persists in all Proton versions I have tried (4.2.7, 3.16.4, 3.7.8), although the intro cut-scene is fine in 3.7.8 (but not the other cut-scenes). The rest of the game runs flawlessly, including in-game cut-scenes. My guess would be that it's related to pulseaudio, but I can't be sure. Any recommendations/ideas on how I can fix this or debug this further?
Edit: Seems to have been fixed in 4.2.9.

NNewbytee 2019-07-02 github

I have a very bizarre issue with the game: after playing the game for a while, could be anything between 1 hour or 5 hours, the game starts getting very choppy. It progressively gets worse over time, to the point where the game almost freezes and weird bugs can appear, such as items disappearing when I move them around in the inventory. At first I thought this was due to the VRAM situation with the GTX 970, but then I realised something: it only happens when I move my mouse (a Logitech G502). No stutter occurs if I move around with my keyboard. Even more odd is that the game is also perfectly smooth if I use my stone age Compaq PS/2 mouse to look around the screen.

To clarify: What occurs is that after playing for a while, when I move my Logitech G502 around, the game at first mostly stutters, but progressively gets worse and begins to temporarily freeze. This occurs regardless of where in the game I do it; running around Novigrad, fighting monsters in a cave, playing Gwent, sorting my inventory: it's all the same, freezes and stutter. Navigating with my keyboard does not produce any stutter and the game plays like it should, and the same goes for my old Compaq PS/2 mouse.

I have not tested other USB mice yet (which might be worth doing).

System info:
GTX 970
nvidia 430.26
Fedora 30 Workstation
Linux 5.1.15

Note that it has occurred with older versions of both the nvidia blob and Linux.

Llieff 2019-07-02 github

@Newbie13XD What happens if you disconnect Logitech G502 and connect it again while stutter occurs?

KKarloKick 2019-07-02 github

Witcher 3 on ubuntu with Steam play won't start

Issue transferred from #2817.
@milka123 posted on 2019-06-17T15:52:24:

Compatibility Report

* Name of the game with compatibility issues: Witcher 3: Wild hunt

* Steam AppID of the game: 292030

System Information

* GPU: NVIDIA GeForce GTX 1050 Ti

* Driver/LLVM version: 396.54, also tried with 418.56

* Kernel version: 4.15.0-51-generic

* Link to full system information report as [Gist](https://gist.github.com/):
  https://gist.github.com/milka123/f9d06e849c2e0e7fed695af72af5d3ef

* Proton version:

I confirm:

* [x]  that I haven't found an existing compatibility report for this game.

* [x]  that I have checked whether there are updates for my system available.

Symptoms

When I start the game, Steam client says it's running but nothing shows up. When I check system monitor I can see witcher3.exe in processes as running but nothing happens.
I couldn't find any more details about potential crashes in steam app log file.

Reproduction

Downloaded Witcher 3: Wild hunt through steam client and tried to run it

steam-292030.log

I have the same issue. Log File is attached
steam-292030_c.log

NNewbytee 2019-07-02 github

@Newbie13XD What happens if you disconnect Logitech G502 and connect it again while stutter occurs?

Can confirm that it changes nothing. I even tried unplugging it while the game was frozen, still took its usual time to unfreeze.

NNewbytee 2019-07-02 github

Update: Tried plugging the G502 into a different USB port, issue persists. I also tried plugging in some old USB Dell mouse I have, issue does not occur with it.

Side note: This is the original G502, not the RGB version.

Ggrigi 2019-07-02 github

Could it be related to this: USB report rate: 1000 Hz (1ms)
Whereas a regular USB mouse is 200Hz?
And PS/2 is 40Hz?

NNewbytee 2019-07-02 github

Could it be related to this: USB report rate: 1000 Hz (1ms)
Whereas a regular USB mouse is 200Hz?
And PS/2 is 40Hz?

Yes, that could very much be it. Nice catch! The G502 is set to use 1000Hz. (or so I think, anyway. I set it to that while using Windows, and I think it is stored in the mouse itself. Not entirely sure how to change this in Linux)

NNewbytee 2019-07-02 github

Also, unrelated issue: Vimme Vivaldi's (the guy who has the place where you can exchange florens and orens for crowns in Novigrad) arms are invisible (as in, no texture, can be seen through) for me with nvidia 430.26 on Proton 4.2-9 (was the same in prior versions of the nvidia blob and Proton too).

Ttriffid 2019-07-02 github

I'm also suddenly having the game not draw anything or show any window
recently, just says 'running' in steam and an idle witcher3.exe process in
process list.

Used to work fine, haven't specifically changed anything, but steam keeps
shoving updates down my throat so maybe it's related to a proton update?

Strace says it's waiting on a futex, but I can't seem to start it under
strace so can't see which futex it might be.

On Tue., 2 Jul. 2019, 20:17 Newbyte, [email protected] wrote:

Could it be related to this: USB report rate: 1000 Hz (1ms)
Whereas a regular USB mouse is 200Hz?
And PS/2 is 40Hz?

Yes, that could very much be it. The G502 is set to use 1000Hz. (or so I
think, anyway. I set it to that while using Windows, and I think it is
stored in the mouse itself. Not entirely sure how to change this in Linux)

—
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub
https://github.com/ValveSoftware/Proton/issues/607?email_source=notifications&email_token=AABICGWVSEU2QYR6FD7ECWDP5NBPHA5CNFSM4FRSLUZ2YY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGODZBCKLI#issuecomment-507651373,
or mute the thread
https://github.com/notifications/unsubscribe-auth/AABICGXHVBL66556J3CFVSTP5NBPHANCNFSM4FRSLUZQ
.

Ooenhan 2019-07-06 github

The same problem with nvidia 430.26 on Proton 4.2-9: witcher3 is running but show any window. Enable PROTON_USE_WINED3D in user_setting.py can make witcher3 work with poor graphic performace, so DXVK maybe the troublemaker.
I remember the earlier version of Proton works fine, but I don't know how to downgrade it to test in steam, anyone can tell me, Thanks

KKarloKick 2019-07-06 github

Disabling precaching the shaders and restarting steam client solved this for me.

See https://steamcommunity.com/app/221410/discussions/0/1640915206462889040/ for the original solution.

Ttriffid 2019-07-06 github

Interesting, turning off shader pre-cache gets the game running for me too!
Sweet

On Sat, 6 Jul 2019 at 14:34, KarloKick [email protected] wrote:

Disabling precaching the shaders and restarting steam client solved this
for me.

—
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub
https://github.com/ValveSoftware/Proton/issues/607?email_source=notifications&email_token=AABICGRFTNZELXEBT7M5HNTP6A4HXA5CNFSM4FRSLUZ2YY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGODZKTS5I#issuecomment-508901749,
or mute the thread
https://github.com/notifications/unsubscribe-auth/AABICGW2FPF5TB77ZP42CX3P6A4HXANCNFSM4FRSLUZQ
.

Kkisak-valve maintainer 2019-07-31 github

Performance regression on The Witcher 3, unstable frame times after 4.11-1

Issue transferred from https://github.com/ValveSoftware/Proton/issues/2921.
@MaCroX95 posted on 2019-07-31T00:26:09:

Compatibility Report

  • Name of the game with compatibility issues: The Witcher 3
  • Steam AppID of the game: 292030

System Information

  • GPU: Nvidia GTX 970
  • Driver/LLVM version: Nvidia 430
  • Kernel version: 5.0.0-21
  • Link to full system information report as Gist: (will update tomorrow, not on my main PC now)
  • Proton version: 4.11-1

I confirm:

  • [X] that I haven't found an existing compatibility report for this game.
  • [X] that I have checked whether there are updates for my system available.

Symptoms

Frame times on The Witcher 3 under same GPU driver, same graphical settings and in the same area are unstable in 4.11-1 when they were completely stable and well performing in 4.2-9. I attach a picture of DXVK_HUD showing frame times in both proton version on same location and behavior in the game.
Link to the image comparing the 2 versions
I was not sure whether the issue is Wine, DXVK or Proton related so I made an issue here and hopefully if it's not a proton issue directly, someone can point it to right direction.

Reproduction

Either open the game under one version or another and compare the performance fps and frame times, stuttering on the latest version is VERY noticable, and isn't related to shader compiling since I've been running around the same location for more than 15 minutes in game.

Ddoitsujin maintainer 2019-07-31 github

@MaCroX95 do you have Vsync enabled by any chance?

I cannot reproduce the problem, but some people have been complaining about Vsync causing severe stutter on some Nvidia systems lately.

Lleillo1975 2019-07-31 github

Performance regression on The Witcher 3, unstable frame times after 4.11-1

Issue transferred from #2921.
@MaCroX95 posted on 2019-07-31T00:26:09:

Compatibility Report

* Name of the game with compatibility issues: The Witcher 3

* Steam AppID of the game: 292030

System Information

* GPU: Nvidia GTX 970

* Driver/LLVM version: Nvidia 430

* Kernel version: 5.0.0-21

* Link to full system information report as [Gist](https://gist.github.com/): (will update tomorrow, not on my main PC now)

* Proton version: 4.11-1

I confirm:

* [x]  that I haven't found an existing compatibility report for this game.

* [x]  that I have checked whether there are updates for my system available.

Symptoms

Frame times on The Witcher 3 under same GPU driver, same graphical settings and in the same area are unstable in 4.11-1 when they were completely stable and well performing in 4.2-9. I attach a picture of DXVK_HUD showing frame times in both proton version on same location and behavior in the game.
Link to the image comparing the 2 versions
I was not sure whether the issue is Wine, DXVK or Proton related so I made an issue here and hopefully if it's not a proton issue directly, someone can point it to right direction.

Reproduction

Either open the game under one version or another and compare the performance fps and frame times, stuttering on the latest version is VERY noticable, and isn't related to shader compiling since I've been running around the same location for more than 15 minutes in game.

I Have the same problem

?ghost 2019-08-01 · hidden on GitHub github

Performance regression on The Witcher 3, unstable frame times after 4.11-1

Issue transferred from #2921.
@MaCroX95 posted on 2019-07-31T00:26:09:

Compatibility Report

* Name of the game with compatibility issues: The Witcher 3

* Steam AppID of the game: 292030

System Information

* GPU: Nvidia GTX 970

* Driver/LLVM version: Nvidia 430

* Kernel version: 5.0.0-21

* Link to full system information report as [Gist](https://gist.github.com/): (will update tomorrow, not on my main PC now)

* Proton version: 4.11-1

I confirm:

* [x]  that I haven't found an existing compatibility report for this game.

* [x]  that I have checked whether there are updates for my system available.

Symptoms

Frame times on The Witcher 3 under same GPU driver, same graphical settings and in the same area are unstable in 4.11-1 when they were completely stable and well performing in 4.2-9. I attach a picture of DXVK_HUD showing frame times in both proton version on same location and behavior in the game.
Link to the image comparing the 2 versions
I was not sure whether the issue is Wine, DXVK or Proton related so I made an issue here and hopefully if it's not a proton issue directly, someone can point it to right direction.

Reproduction

Either open the game under one version or another and compare the performance fps and frame times, stuttering on the latest version is VERY noticable, and isn't related to shader compiling since I've been running around the same location for more than 15 minutes in game.

Update on the issue:

  1. I thought that the issue would be caused by Proton 4.11-1 being placed to my HDD instead of SSD, that turned out not to be the case, problem still persists after having it installed on my SSD.

Conclusion 1: Changed behavior that Proton is now placed on default game library (HDD in my case) instead of the location of where the Steam is installed has no effect on the issue therefore they're not related.

  1. DXVK developer asked me to test Proton 4.11-1 with DXVK version from 4.2-9 and the other way around. Here is the result:

Screenshot of tests

Conclusion 2: The problem is not related to updated DXVK library.

Remaining options are either Wine regression or Proton regression.

Kkisak-valve maintainer 2019-08-01 · hidden on GitHub github

Hello @MaCroX95, please add PROTON_LOG=1 %command% to the game's launch options and drag and drop the generated $HOME/steam-$APPID.log into the comment box.

?ghost 2019-08-01 · hidden on GitHub github

Hello @MaCroX95, please add PROTON_LOG=1 %command% to the game's launch options and drag and drop the generated $HOME/steam-$APPID.log into the comment box.

@kisak-valve Here you are sir:
gist to logs (Proton 4.2-9 vs Proton 4.11-1)

Ssirkon 2019-08-02 github

It became much, I mean MUCH smoother for me.

NNewbytee 2019-08-03 github

It became much, I mean MUCH smoother for me.

With fsync, or just with Proton 4.11-1?

Ssirkon 2019-08-03 github

@Newbie13XD with Proton 4.11-1, vSync always was on

1080Ti, nVidia 430

Hierarch Square in Novigrad always was problematic with Proton: the FPS was clearly below 60 while in other areas it was lower than 120 (less than two in each 1/60 of a second to be precise). So, it is above 60 FPS in that square and above 120 in pretty much every place.

NNewbytee 2019-08-03 github

@Newbie13XD with Proton 4.11-1, vSync always was on

No no, not v-sync. I'm referring to what Valve speaks of here: https://github.com/ValveSoftware/Proton/releases/tag/proton-4.11-1b

Specifically, the third bullet.

1080Ti, nVidia 430

Hierarch Square in Novigrad always was problematic with Proton: the FPS was clearly below 60 while in other areas it was lower than 120 (less than two in each 1/60 of a second to be precise). So, it is above 60 FPS in that square and above 120 in pretty much every place.

Great to hear that!

Ssirkon 2019-08-03 github

@Newbie13XD as I am using VSync I don't really have more than 60FPS, I meant sampling theorem and its consequences.

PS sorry to hear others having issues, hope they will see this improvement soon.

?ghost 2019-08-08 · hidden on GitHub github

@kisak-valve the issue has been resolved for me in 4.11-2, still not sure what was causing the unstable frame times, and I don't see any specific "fixes" in the changelog that could have solved those issues, but just though I might let you know that the performance is now back on par with 4.2-9.

Ssirkon 2019-08-14 github

Meanwhile 4.11-2 performs much worse for me, it is basically a return to 4.2-9 level of performance.

NNewbytee 2019-08-14 github

Meanwhile 4.11-2 performs much worse for me, it is basically a return to 4.2-9 level of performance.

Probably due to some reverted changes for NVIDIA cards in DXVK.

NNewbytee 2019-08-15 github

Compatibility report

  • Name of the game with compatibility issues: The Witcher 3: Wild Hunt
  • Steam AppID of the game: 292030

System info

steam-292030.log

Symptoms

Game randomly freezes, mostly when getting on Roach. It takes the entire PC down with it, and requires a forced reset to become responsive again.

This has yet to occur with any other games (Minecraft, The Awesome Adventures of Captain Spirit, Sims 3, and Sims 4 tested).

Reproduction

Play the game and eventually experience freeze. No freezes like these ever occurred on my other PC however.

TTheGreatestJannet 2019-09-08 github

I am also experiencing the stuttering and high frame times on Proton 4.11-3. Downgrading to Proton 4.2-9 solves the issue. I have a GTX 1060 6GB with the 430.40 drivers.

Kkisak-valve maintainer 2019-11-02 github

The Witcher 3 and Mesa 19.3/git (292030)

Issue transferred from https://github.com/ValveSoftware/Proton/issues/3185.
@atomsymbol posted on 2019-11-02T15:47:02:

Compatibility Report

  • Name of the game with compatibility issues: The Witcher 3
  • Steam AppID of the game: 292030

System Information

I confirm:

  • [x] that I haven't found an existing compatibility report for this game.
  • [x] that I have checked whether there are updates for my system available.

Symptoms

The Witcher 3 works fine with Mesa 19.2.2, but fails to start with Mesa git. This issue may also be reproducible with other Proton games.

The game seems to start fine in plain Wine 4.18 outside of Steam.

Reproduction

  1. Update Mesa to latest git version
  2. Start The Witcher 3
  3. An exception is reported by Proton's Wine in the terminal running Steam. The game does not create any window.
?ghost 2019-11-03 github

@kisak-valve wrote:

The conclusion should be that DXVK in Proton goes to a vulkan render path instead of OpenGL with upstream wine's wined3d.

Getting a backtrace from winedbg is very slow, but it pointed me to the fact that I am compiling LLVM and Mesa with static libstdc++ on my machine. I recompiled the packages with shared libstdc++ and the issue has been resolved.

Kkisak-valve maintainer 2019-11-10 github

The Witcher 3 + Mesa 19.1 - game trying to start, but windows hangs (black screen)

Issue transferred from https://github.com/ValveSoftware/Proton/issues/3203.
@Gman1988 posted on 2019-11-10T19:58:41:

Compatibility Report

  • Name of the game with compatibility issues: The Witcher 3
  • Steam AppID of the game: 292030

System Information

I confirm:

  • [ x ] that I haven't found an existing compatibility report for this game.
  • [ x ] that I have checked whether there are updates for my system available.

steam-292030.log

Symptoms

Gamer is trying to start, new full-screen window opens, but stays black. After a while it stops responding.

Reproduction

Just start the game.

Kkisak-valve maintainer 2019-11-10 github

Hello @Gman1988, your system information hints that you're trying to run the game with your Intel Haswell gpu, which has experimental vulkan support. It may be worth giving https://github.com/ValveSoftware/Proton/wiki/For-AMD-users-having-issues-with-non-OpenGL-games a read. You have a Sea Island (CIK) video card, and the radeon kernel module does not support Vulkan.

4414n 2019-12-24 github

I just bought and installed Witcher 3 GOTY edition and found out that it wouldn't work out of the box on my system with latest Proton-4.11 (Steam system info): by pushing the play button in Steam, a number of witcher3.exe processes would be launched, but nothing would show up even after some time.

By reading https://github.com/ValveSoftware/Proton/issues/607#issuecomment-508901749 I deleted everything under ~/.steam/steam/SteamApps/shadercache/292030/ and BAM, the game started successfully.

Ttriffid 2020-02-21 github

After recent updates, Witcher 3 won't start

steam-witcher3-crash.txt - critical line seems to be wine: Unhandled page fault on read access to 0000000000000000 at address 00007FB948436C85 (thread 0037)...

Tried with both proton-3.16-9 and proton-5.0-2. Trying proton-4.11-12 now.

I have deleted the shader pre-cache for 292030 and turned it off in options.

Using nvidia-440.59 on a GTX1060, game previously ran perfectly.

4414n 2020-03-07 github

@triffid no problem here even with latest Proton versions...
Are you still experiencing those issues?

Ttriffid 2020-03-08 github

It seems that there's an incompatibility between winepulse and apulse - see https://github.com/i-rinat/apulse/issues/86

Turns out it's not related to Witcher 3 at all, but I'd appreciate suggestions on the appropriate spot to report this issue.

Other non-steam windows applications on my system weren't affected because I choose to build wine without pulseaudio support.

Linux native steam applications also seem to be unaffected by my use of apulse.

So on my system, only Proton is affected because I don't get to choose which audio pathways it thinks it can use.

Ideally, the winepulse/apulse issue should be fixed.

Failing that, Proton should detect apulse and fall back to alsa, like what happens if there's no pulse library available at all.

4414n 2020-03-08 github

@triffid have you tried forcing ALSA output from Proton using winetricks sound=alsa in the relevant Proton/wine prefix?
If you'd like, I'm working on a simple shell-based tool to make these things easy.

Ttriffid 2020-03-09 github

Ooh that sounds handy, I'll try it, thanks!

4414n 2020-03-09 github

Ooh that sounds handy, I'll try it, thanks!

Here you go: https://github.com/414n/protonsh

When you launch it, it will ask what Proton prefix and Proton version you want to use, then it will launch another shell instance with the proper environment overrides to execute winetricks tasks and the like inside that prefix.

Nnutta-git 2020-03-11 github

Proton 5.0.4
Fedora31 Workstation
5.6.0-0.rc5.git0.1.fc31.x86_64
Mesa 20.0.0 Devel

Problem:
Mouse Cursor does not work properly in menus.
Can't select game options like Resume, Options, etc... using mouse cursor.

Can someone else verify?
https://github.com/nutta-git/logs-/blob/master/steam-292030(The Witcher 3).log

NNewbytee 2020-03-11 github

Proton 5.0.4
Fedora31 Workstation
5.6.0-0.rc5.git0.1.fc31.x86_64
Mesa 20.0.0 Devel

Problem:
Mouse Cursor does not work properly in menus.
Can't select game options like Resume, Options, etc... using mouse cursor.

Can someone else verify?
https://github.com/nutta-git/logs-/blob/master/steam-292030(The Witcher 3).log

It's fine for me. Maybe a kernel or Mesa bug?

Proton 5.0-4
Fedora 31 Workstation
5.5.8-1.fc31
nvidia 440.64

AAsciiWolf 2020-03-11 github

This may also be a Wayland issue. Nvidia does not use Wayland on Fedora.

Nnutta-git 2020-03-11 github

Mouse seems to work fine with TKG's Proton.

Nnutta-git 2020-03-13 github

Proton 5.0.4
Fedora31 Workstation
5.6.0-0.rc5.git0.1.fc31.x86_64
Mesa 20.0.0 Devel

Problem:
Mouse Cursor does not work properly in menus.
Can't select game options like Resume, Options, etc... using mouse cursor.

Can someone else verify?
https://github.com/nutta-git/logs-/blob/master/steam-292030(The Witcher 3).log

Seem like a Wayland Issue, game works well with X-Org.
problem solved for now

NNewbytee 2020-03-13 github

Proton 5.0.4
Fedora31 Workstation
5.6.0-0.rc5.git0.1.fc31.x86_64
Mesa 20.0.0 Devel
Problem:
Mouse Cursor does not work properly in menus.
Can't select game options like Resume, Options, etc... using mouse cursor.
Can someone else verify?
https://github.com/nutta-git/logs-/blob/master/steam-292030(The Witcher 3).log

Seem like a Wayland Issue, game works well with X-Org.
problem solved for now

Out of curiosity, does it work in (X)Wayland when using older versions of Proton?

Nnutta-git 2020-03-13 github

Proton 5.0.4
Fedora31 Workstation
5.6.0-0.rc5.git0.1.fc31.x86_64
Mesa 20.0.0 Devel
Problem:
Mouse Cursor does not work properly in menus.
Can't select game options like Resume, Options, etc... using mouse cursor.
Can someone else verify?
https://github.com/nutta-git/logs-/blob/master/steam-292030(The Witcher 3).log

Seem like a Wayland Issue, game works well with X-Org.
problem solved for now

Out of curiosity, does it work in (X)Wayland when using older versions of Proton?

same problem happen with proton 4.11.latest with wayland

Ssoredake 2020-03-18 · hidden on GitHub github

Proton 5.0-4, overlay not working.

steam-292030.log

Ssoredake 2020-03-20 · hidden on GitHub github

False alarm, overlay is working, just binding is switched from shift+del to default shift+tab for some reason.

?ghost 2020-03-22 github

Proton 5.0.4
Fedora31 Workstation
5.6.0-0.rc5.git0.1.fc31.x86_64
Mesa 20.0.0 Devel

Problem:
Mouse Cursor does not work properly in menus.
Can't select game options like Resume, Options, etc... using mouse cursor.

Can someone else verify?
https://github.com/nutta-git/logs-/blob/master/steam-292030(The Witcher 3).log

I fixed this issue by disabling hardware cursor. Running latest sway and arch linux as of today.

Nnutta-git 2020-03-23 github

@billop Disabling Hardware Cursor has fixed this issue.
Thank you!

Kkisak-valve maintainer 2020-03-23 github

CTD on start game on Navi10

Issue transferred from https://github.com/ValveSoftware/Proton/issues/3682.
@Atrosha posted on 2020-03-23T17:46:49:

Compatibility Report

  • Name of the game with compatibility issues: Witcher 3
  • Steam AppID of the game: 292030

System Information

I confirm:

  • [x] that I haven't found an existing compatibility report for this game.
  • [x] that I have checked whether there are updates for my system available.

steam-292030.log
(MESA_DEBUG)terminal,log

Symptoms

Reproduction

Kkisak-valve maintainer 2020-03-23 github

Hello @Atrosha, wine: /mnt/storage/LinuxGames/steamapps/compatdata/292030/pfx is not owned by you is the line of interest from your log.

I'm going to take a guess that /mnt/storage is an NTFS partition, which is known to be finicky with Proton. If that is true, please give https://github.com/ValveSoftware/Proton/wiki/Proton-FAQ#none-of-my-games-are-launching--prefix-not-owned-by-you-error / https://github.com/ValveSoftware/Proton/wiki/Using-a-NTFS-disk-with-Linux-and-Windows a read.

AAtrosha 2020-03-23 github

Thanks @kisak-valve this helps a bit. Also I moved my Proton installation to home folder. But now the game starts at Steam and looks like running but nothing happens. And only force exit from Steam helps - the Stop button doesn't work.

I'm using Fedora Linux Rawhide (32)

Maybe this helps a bit?

[atrosha@localhost ~]$ cd ~/.local/share/Steam/ubuntu12_32
[atrosha@localhost ubuntu12_32]$ LD_LIBRARY_PATH=".:${LD_LIBRARY_PATH}" ldd $(file *|sed '/ELF/!d;s/:.*//g')|grep 'not found'|sort|uniq
	libbz2.so.1.0 => not found
	libdbus-glib-1.so.2 => not found
	libnm-glib.so.4 => not found
	libnm-util.so.2 => not found
	libudev.so.0 => not found

I found it here: https://github.com/ValveSoftware/steam-for-linux/issues/3959#issuecomment-311404926

I tried to sudo dnf install NetworkManager-libnm-devel.i686 with no result, because libnm-glib.so.4 and libnm-util.so.2 isn't there.

4414n 2020-03-23 github

@Atrosha maybe you're having the same issue as me, namely a bogus shader cache: https://github.com/ValveSoftware/Proton/issues/607#issuecomment-568767767

AAtrosha 2020-03-23 github

Thanks @414n but that isn't it. Also I found this game specific by launching Thea 2 (Unity engine). Thea 2 works fine. And Hellblade (UE4)

When I start The Witcher 3 "prepare to launch" dialogue stage takes too long about 4-5 minutes but installs nothing.

Tttshaw1 2020-04-12 · hidden on GitHub github

Witcher 3 on ubuntu with Steam play won't start

Issue transferred from #2817.
@milka123 posted on 2019-06-17T15:52:24:

Compatibility Report

* Name of the game with compatibility issues: Witcher 3: Wild hunt

* Steam AppID of the game: 292030

System Information

* GPU: NVIDIA GeForce GTX 1050 Ti

* Driver/LLVM version: 396.54, also tried with 418.56

* Kernel version: 4.15.0-51-generic

* Link to full system information report as [Gist](https://gist.github.com/):
  https://gist.github.com/milka123/f9d06e849c2e0e7fed695af72af5d3ef

* Proton version:

I confirm:

* [x]  that I haven't found an existing compatibility report for this game.

* [x]  that I have checked whether there are updates for my system available.

Symptoms

When I start the game, Steam client says it's running but nothing shows up. When I check system monitor I can see witcher3.exe in processes as running but nothing happens.
I couldn't find any more details about potential crashes in steam app log file.

Reproduction

Downloaded Witcher 3: Wild hunt through steam client and tried to run it

steam-292030.log

I believe I have the same symptoms. One thing to note is that a few seconds after starting the game the process closes and the button goes for "Cancel" back to "Play." I've disabled shader pre-caching and deleted the files. Tried with and without forcing wined3d in the launch options.

OpenGL version string: 4.5 (Compatibility Profile) Mesa 19.2.8

26:00.0 VGA compatible controller: Advanced Micro Devices, Inc. [AMD/ATI] Ellesmere [Radeon RX 470/480/570/570X/580/580X] (rev c7) Subsystem: Micro-Star International Co., Ltd. [MSI] Radeon RX 480 Kernel driver in use: amdgpu Kernel modules: amdgpu
Kernel version 5.3.0-46

Proton logs attached.
steam-292030_normal.log
steam-292030_wind3d.log

Tttshaw1 2020-04-13 · hidden on GitHub github

Witcher 3 on ubuntu with Steam play won't start
Issue transferred from #2817.
@milka123 posted on 2019-06-17T15:52:24:

Compatibility Report

* Name of the game with compatibility issues: Witcher 3: Wild hunt

* Steam AppID of the game: 292030

System Information

* GPU: NVIDIA GeForce GTX 1050 Ti

* Driver/LLVM version: 396.54, also tried with 418.56

* Kernel version: 4.15.0-51-generic

* Link to full system information report as [Gist](https://gist.github.com/):
  https://gist.github.com/milka123/f9d06e849c2e0e7fed695af72af5d3ef

* Proton version:

I confirm:

* [x]  that I haven't found an existing compatibility report for this game.

* [x]  that I have checked whether there are updates for my system available.

Symptoms

When I start the game, Steam client says it's running but nothing shows up. When I check system monitor I can see witcher3.exe in processes as running but nothing happens.
I couldn't find any more details about potential crashes in steam app log file.

Reproduction

Downloaded Witcher 3: Wild hunt through steam client and tried to run it
steam-292030.log

I believe I have the same symptoms. One thing to note is that a few seconds after starting the game the process closes and the button goes for "Cancel" back to "Play." I've disabled shader pre-caching and deleted the files. Tried with and without forcing wined3d in the launch options.

OpenGL version string: 4.5 (Compatibility Profile) Mesa 19.2.8

26:00.0 VGA compatible controller: Advanced Micro Devices, Inc. [AMD/ATI] Ellesmere [Radeon RX 470/480/570/570X/580/580X] (rev c7) Subsystem: Micro-Star International Co., Ltd. [MSI] Radeon RX 480 Kernel driver in use: amdgpu Kernel modules: amdgpu
Kernel version 5.3.0-46

Proton logs attached.
steam-292030_normal.log
steam-292030_wind3d.log

Solved by installing vulkan

AAtrosha 2020-04-14 github

Meet problem with the same symptoms and Vulkan already installed.

After reinstalling the game the installation of redistributables stuck at step 3.

It was Fedora and after newest updates all works perfectly.

SStatusCode404 2020-04-14 · hidden on GitHub github

Hi Proton experts!
Witcher 3 won't start at all with 5.0-5 but works with previous version 4.11-13.
Looks like something broke from 4.11 to 5.0.
Any help would be appreciated! Thanks in advance!

Proton: 1584918606 proton-5.0-5b
OS: Ubuntu 18.04.4 (kernel: 5.5.16-050516-generic)
Renderer : AMD Radeon R9 200 Series (HAWAII, DRM 3.36.0, 5.5.16-050516-generic, LLVM 9.0.0)
Version : 4.5 (Compatibility Profile) Mesa 19.2.8
Vulkan: mesa-vulkan-drivers is already the newest version (19.2.8-0ubuntu0~18.04.3).
Vulkan32: mesa-vulkan-drivers:i386 is already the newest version (19.2.8-0ubuntu0~18.04.3)
LOG: steam-292030.log

Ssoredake 2020-04-14 · hidden on GitHub github

@StatusCode404 try with next version, and try to delete your prefix
image

SStatusCode404 2020-04-14 · hidden on GitHub github

Not sure if I understood your instructions correctly.

  1. In Steam library I right clicked on "Pronton 5.0" under the horizontal scroll frame "Recent Games"
  2. I then clicked on "Properties" > "BETAS"
  3. Then selected "next -"
  4. "Close"
  5. Then on the same horizontal scroll frame I right clicked on Witcher 3 > "Properties" > "General"
  6. Chose proton 5.0-5

I tried to execute and run the game after the 6 points above and no joy. Same "Fatal error".
Works ok except for pauses and stutters on 4.11-13

NNewbytee 2020-04-14 · hidden on GitHub github

Not sure if I understood your instructions correctly.

1. In Steam library I right clicked on "Pronton 5.0" under the horizontal scroll frame "Recent Games"

2. I then clicked on "Properties" > "BETAS"

3. Then selected "next -"

4. "Close"

5. Then on the same horizontal scroll frame I right clicked on Witcher 3 > "Properties" > "General"

6. Chose proton 5.0-5

I tried to execute and run the game after the 6 points above and no joy. Same "Fatal error".
Works ok except for pauses and stutters on 4.11-13

Can you send the log? Add PROTON_LOG %command% to the game's launch options and upload the file steam-292030.log that will be created in your home directory here.

SStatusCode404 2020-04-14 · hidden on GitHub github

@Newbytee I had already attached it to my initial post 4 spots above... https://github.com/ValveSoftware/Proton/issues/607#issuecomment-613333539

SStatusCode404 2020-04-16 · hidden on GitHub github

Proton 5.0-6 Test
Just an update on my previous comments...
LOG: steam-292030.log

I tried the Witcher 3 on the just release 5.0-6 with mesa 19.2.8 with R9-290X and same failure and effect as 5.0-5.
Looks like regression was NOT fixed.

Works OK with occasional stutter with proton 4.11-13, mesa 19.2.8 with R9-290X.

FFatmice 2020-04-16 github

I would like some help .... My game launches with the wrongly detected resolution. This has happened to other games too...

Screenshot from 2020-04-15 19-43-26

As you can see, I have two physical displays; (1) is a 1920x1080 and (2) is 2560x1440

I launched the game with DXVK_FILTER_DEVICE_NAME="VEGA20" gamemoderun %command%
The game then launches on display (2) at the upper left corner and is using XWAYLAND.

Report from xwininfo

xwininfo: Window id: 0x320000b "The Witcher 3"

  Absolute upper-left X:  0
  Absolute upper-left Y:  27
  Relative upper-left X:  0
  Relative upper-left Y:  27
  Width: 1920
  Height: 1080
  Depth: 24
  Visual: 0x58
  Visual Class: TrueColor
  Border width: 0
  Class: InputOutput
  Colormap: 0x3000001 (not installed)
  Bit Gravity State: NorthWestGravity
  Window Gravity State: NorthWestGravity
  Backing Store State: NotUseful
  Save Under State: no
  Map State: IsViewable
  Override Redirect State: no
  Corners:  +0+27  -2560+27  -2560-333  +0-333
  -geometry 1920x1080+0+27


As you can see the geometry is wrong...

Report from xprop

xprop -id 0x320000b
WM_STATE(WM_STATE):
		window state: Normal
		icon window: 0x0
_NET_WM_DESKTOP(CARDINAL) = 0
_GTK_EDGE_CONSTRAINTS(CARDINAL) = 170
_NET_WM_ALLOWED_ACTIONS(ATOM) = _NET_WM_ACTION_MOVE, _NET_WM_ACTION_CHANGE_DESKTOP, _NET_WM_ACTION_CLOSE, _NET_WM_ACTION_ABOVE, _NET_WM_ACTION_BELOW
_NET_WM_BYPASS_COMPOSITOR(CARDINAL) = 0
_NET_WM_STATE(ATOM) = 
_NET_WM_ICON(CARDINAL) = 	Icon (32 x 32):
	                                
	                                
	                                
	    ▒    ░            ░    ▒░
	   ░▓▒   ▒░   ░      ▒▓   ▒▓▒
	    █▓▒▒ ▓▓ ▒▒░░░░▒▒░▓▓░▒▒▓█░
	    ▓█▓▒░░▒░░▒░░░░▒░▒▓▒░▒▓▓▓
	    ░▓▓▓▒░░▒░ ░  ░  ▒▒▒▒▓▓▓▒
	     ▒▓▓▓▓░░░  ░    ▒░▓▓▒▓▓
	    ░▓▓▒▒▒░░ ░░░░░   ░▒▒▒▓▓░░
	  ░▓▓▓██░ ▓▒  ░░░░  ░▒░░▓█▓▓▓░
	    ▓▓▓▒▒░░░░▒░░░░▒░░░░░▓▓▓▒
	    ▒▓▒░░░░ ░▒▒▒░▒▒▒ ░░░▒▒▓▒
	 ░▓▓▒▒▒▒░  ░░░░░░░░░ ░  ░▒▒▒▓▓▒ 
	 ▓▓▓▓▓▓▓▓▓▓▓▓▒▒  ░▒▓▓▓▓▓▓▓▓▓▓▓▓░
	   ░▒▓▓███▓░▓▓▓▓▓▓▓▓ ▒███▓▓▒░
	      ▓▓▓▒ ░▓▓▓▓▓▓▓▓░ ░▓▓▓
	     ▒█▓░░▓██▓▒▓▓▒▓██▒░░▓█▒
	    ░▓▓░▒▓███▓▓▓▓▓▓███▓▒░▓█░
	    ▓▓▒▒███▓▓██▓▓██▓▓███▓░▓▓
	   ▒▓▓▓█▓▓▓▒▓██████▓▒▓█▓█▓▓▓▒
	  ░▓▓█▓▒░▓▒▒█▓▓▓▓▓▓█▓▒▓░▒▓█▓▓░
	 ░▓▓▒░   ▓▒██▓▓▓▓▓▓██▒▒░  ░▒▓▓░
	 ░░     ░▒▓▓▓▓▓▓▓▓▓▓█▓░▒     ░░
	        ▒▓█▓ ░▓▓▓▓▒ ▒█▓▒
	        ▓▓░  ░▓▓█▓░  ░▓▓
	       ░▓    ░▒▓▓▒░    ▓░
	              ▒▒▓▒░
	              ▓▓▓▓
	              ░░░░
	                                
	                                


_NET_WM_NAME(UTF8_STRING) = "The Witcher 3"
WM_ICON_NAME(STRING) = "The Witcher 3"
WM_NAME(STRING) = "The Witcher 3"
WM_HINTS(WM_HINTS):
		Client accepts input or input focus: True
		Initial state is Normal State.
		bitmap id # to use for icon: 0x30000a5
		bitmap id # of mask for icon: 0x30000a7
		window id # of group leader: 0x320000b
_NET_WM_WINDOW_TYPE(ATOM) = _NET_WM_WINDOW_TYPE_NORMAL
_MOTIF_WM_HINTS(_MOTIF_WM_HINTS) = 0x3, 0x24, 0x0, 0x0, 0x0
WM_NORMAL_HINTS(WM_SIZE_HINTS):
		program specified location: 0, 0
		program specified minimum size: 1920 by 1080
		program specified maximum size: 1920 by 1080
		window gravity: Static
_NET_WM_USER_TIME_WINDOW(WINDOW): window id # 0x3000016
XdndAware(ATOM) = BITMAP
_NET_WM_PID(CARDINAL) = 32254
WM_LOCALE_NAME(STRING) = "en_US.UTF-8"
WM_CLIENT_MACHINE(STRING) = "yellowtrain"
WM_CLASS(STRING) = "witcher3.exe", "steam_app_292030"
WM_PROTOCOLS(ATOM): protocols  WM_DELETE_WINDOW, _NET_WM_PING
STEAM_GAME(CARDINAL) = 292030

So clearly the WM_SIZE_HINTS is wrong...

Report from xrandr

[ocelot@yellowtrain ~]$ xrandr --query
Screen 0: minimum 16 x 16, current 4480 x 1440, maximum 32767 x 32767
XWAYLAND0 connected 1920x1080+2560+0 (normal left inverted right x axis y axis) 530mm x 310mm
   1920x1080     59.96*+
   1440x1080     59.99  
   1400x1050     59.98  
   1280x1024     59.89  
   1280x960      59.94  
   1152x864      59.96  
   1024x768      59.92  
   800x600       59.86  
   640x480       59.38  
   320x240       59.52  
   1680x1050     59.95  
   1440x900      59.89  
   1280x800      59.81  
   720x480       59.71  
   640x400       59.95  
   320x200       58.96  
   1600x900      59.95  
   1368x768      59.88  
   1280x720      59.86  
   1024x576      59.90  
   864x486       59.92  
   720x400       59.55  
   640x350       59.77  
XWAYLAND1 connected 2560x1440+0+0 (normal left inverted right x axis y axis) 600mm x 340mm
   2560x1440    143.87*+
   1920x1440    143.84  
   1600x1200    143.89  
   1440x1080    143.80  
   1400x1050    143.89  
   1280x1024    143.79  
   1280x960     143.86  
   1152x864     143.75  
   1024x768     143.66  
   800x600      143.83  
   640x480      143.30  
   320x240      142.05  
   1920x1200    143.89  
   1680x1050    143.88  
   1440x900     143.86  
   1280x800     143.67  
   720x480      143.35  
   640x400      143.37  
   320x200      141.40  
   2048x1152    143.88  
   1920x1080    143.80  
   1600x900     143.81  
   1368x768     143.77  
   1280x720     143.67  
   1024x576     143.91  
   864x486      143.63  
   720x400      143.29  
   640x350      143.57

Clearly, display (2) is capable of 2560x1440...Yet the application is not getting the right hint.

Now I can force the resolution by setting it in user.settings file found under steamuser/My Documents/The Witcher 3/ of the pfx and the game will launch using the entire display (2). However, it still detected the wrong resolution within the game's graphical settings...

[Viewport]
VSync=false
Resolution="2560x1440"
FullScreenMode=1
VSyncThreshold=0
OutputMonitor=0

Thus is there a way to give the game some hints of the right resolution through DXVK's variables or whatever else within the launch options? I've tried DXVK_FILTER_DEVICE_NAME="VEGA20" gamemoderun %command% -w 2560 -h 1440 ... It doesn't work.

Does this have any bearing to the problem? https://gitlab.gnome.org/GNOME/mutter/commit/f241bdba934f938c0a97587806d02c0f2464f221

SShalokShalom 2020-04-28 · hidden on GitHub github

I really really look forward to the day, when a Steam update does not break all the games I have bought.

Screenshot_20200428_113233
Screenshot_20200428_113223

Why is this not accessible? I just played yesterday, sleep mode, awoke it today, and not playable anymore. I had to enter the password upon login, which is normally not the case.

NNewbytee 2020-04-28 · hidden on GitHub github

@ShalokShalom It seems like Proton has been disabled, or at least been disabled for all games. Try re-enabling it in settings?

Ssoredake 2020-04-28 · hidden on GitHub github

I really really look forward to the day, when a Steam update does not break all the games I have bought.

Screenshot_20200428_113233
Screenshot_20200428_113223

Why is this not accessible? I just played yesterday, sleep mode, awoke it today, and not playable anymore. I had to enter the password upon login, which is normally not the case.

image

recheck this option to fix this

SShalokShalom 2020-04-28 · hidden on GitHub github

Yeah, I figured this myself and just came to report about it. This has fixed my issue. Thanks a lot, dear community.

SShalokShalom 2020-04-28 github

Another issue:

Screenshot_20200428_115943

As you can see, is the Controller detected, and yet, it does not function at all.
It is a newly bought Nacon PS4 controller and supposed to be working, as far as I know.

Can anybody help me with that?

EDIT: OK, out of nowhere, the controller started suddenly to function somewhat, while Gerald is now jumping consistently through the game world, with me not pressing any button, and half of the inputs are still not registered. Is this a potentially broken controller?

EDIT 2: It is mainly the left stick, the other buttons and controls seem to work fine.
EDIT 3: The left control stick is working in the Dolphin Emulator and in the Witcher menu.

Ssoredake 2020-04-28 github

Try to disable steam input.

SShalokShalom 2020-04-28 github

Works

EDIT: Of course, it has its own set of issues. A is now used to both sprint and open the menu and sometimes, the share button offers the escape menu and sometimes the quick selection for the signs.

Ddearlygparted 2020-05-01 github

Recently completed my first playthrough (entirely on Linux) and started New Game+. There's this one NPC with invisible arms:
Screenshot from 2020-04-30 19-41-22

No real issues other than that. Performance is great with Feral's GameMode.
Running Arch Linux with 5.6.8 kernel, Ryzen 5 2600, GTX 1060, Nvidia 440.82, Proton 5.0-6.

Ddoitsujin maintainer 2020-05-02 github

@dearlygparted Vivaldi's arms being invisible is a known game bug, happens on Windows as well and apparently even on consoles.

Ffls2018 2020-05-02 github

There's actually a few mods that fix that as well as other graphical issues.

SShalokShalom 2020-05-03 github

Do you have a guideline how to install mods?

Sshmerl 2020-05-03 github

I install mods manually. Sometimes you need to merge scripts.

Sshmerl 2020-05-03 github

Just for the reference, Vivaldi has hands with AMD RX 5700 XT.

NNewbytee 2020-05-04 github

Do you have a guideline how to install mods?

If the mod comes in the form of a DLL you need to do a DLL override in Wine, otherwise it shouldn't be any different from how you do it on Windows.

SShalokShalom 2020-05-06 github

I install mods manually. Sometimes you need to merge scripts.

And how to do this? I have no previous experience with installing mods in a non-Linux native game at all and have no idea where to start. There are also no docs, since that would make things obviously too easy. :D

Thanks

Ssoredake 2020-05-06 github
NNewbytee 2020-05-06 github

I install mods manually. Sometimes you need to merge scripts.

And how to do this? I have no previous experience with installing mods in a non-Linux native game at all and have no idea where to start. There are also no docs, since that would make things obviously too easy. :D

Thanks

I don't think this is the place to discuss modding, but follow guides like on Windows, except drop the files into the relevant subdirectory of Witcher 3's compatdata directory (can be found in your Steam Library directory). You will find a Windows-like file tree there.

SShalokShalom 2020-05-07 github

It might be the place to discuss documentation for such things.

SStatusCode404 2020-05-08 · hidden on GitHub github

Proton 5.0-7 Test
Just an update on my previous comments...
LOG: steam-292030.log

Summary:
I tried the Witcher 3 on the just release 5.0-7 with mesa 20.0.6 Kisak's Valve's PPA with R9-290X and same failure and effect as 5.0-5/6.
Looks like regression was NOT fixed.

Works OK with with proton 4.11-13, mesa 20.0.6 with R9-290X with or without ACO

Ddoitsujin maintainer 2020-05-09 · hidden on GitHub github

@StatusCode404 which regression are you talking about?

NNewbytee 2020-05-09 · hidden on GitHub github

@StatusCode404 which regression are you talking about?

https://github.com/ValveSoftware/Proton/issues/607#issuecomment-613333539

SStatusCode404 2020-05-16 · hidden on GitHub github

Hi All, thanks to @kisak-valve and @HenrikHolst I was able to fix the problem by deleting all the old vulkan json files in /etc/vulkan/icd.d/

If you're stuck in the same situation, you can do the same thing! 5.0-7 now works!

MMollyMaclachlan 2020-05-16 github

Witcher 3 works perfectly on my system right out of the box.

To install mods you download them from Nexus and drop the mod folder (which should be named mod[ModName]) into the mods folder in Witcher 3's compatdata folder. Some of them require the script merger to be compatible with each other; that can also be found on Nexus and as far as I know, works perfectly.

Ssoredake 2020-06-05 github

Found a bug, after some amount of alt-tabs game switches to windowed mode while still being in fullscreen mode according to settings, maybe a game bug but sharing anyway.

To reproduce it fast you need to press Meta+PD then click on witcher in task bar, repeat until success:

https://user-images.githubusercontent.com/5204968/104815053-e936aa80-581a-11eb-8c97-21f48a9cd330.mp4

And log, if it's a wine problem:
steam-292030.log

Ssoredake 2020-06-06 github

Same bug, after alt-tabbing while loading game becomes like this:
Screenshot_20200606_152847

After another alt-tabbing game becomes windowed.

steam-292030.log

Ppredacon 2020-06-11 github

same issue as @pclouds no keyboard or mouse in-game
https://github.com/ValveSoftware/Proton/issues/607#issuecomment-461266351
tested with all available versions of proton in steam native & runtime, tested unloading most controller modules, unloading kwin, tested with another keyboard and mouse, tested forcing off steam input. nothing makes sense.
I'm running archlinux (5.6.15-arch1-1) & KDE5
Run out of ideas for now, any help will be apreciated.

Ppclouds 2020-06-11 github

It works for me now (for probably 6 months). No idea what caused it.

Ssoredake 2020-06-14 github

Whitelist Request

  • Name of the game to be whitelisted: The Witcher 3
  • Steam AppID of the game: 292030

System Information

I confirm:

  • [X] that pressing the Play button in the Steam client is sufficient.
  • [ ] that runtime config options are necessary to run the game.
  • [ ] that no workarounds other than PROTON_* env variables are necessary.

Issues

Ppredacon 2020-06-15 github

same issue as @pclouds no keyboard or mouse in-game
[#607 (comment)](https://github.com/ValveSoftware/Proton/issues/607#issuecomment-461266351)
tested with all available versions of proton in steam native & runtime, tested unloading most controller modules, unloading kwin, tested with another keyboard and mouse, tested forcing off steam input. nothing makes sense.
I'm running archlinux (5.6.15-arch1-1) & KDE5
Run out of ideas for now, any help will be apreciated.

Actually found a fix for this. Under folder "/home/user/.local/share/Steam/" there is two folders with same name but some characters in upper case like that: "SteamApps" and "steamapps".
Deleting the lower case folder "steamapps" solve the problem. Maybe steam could redownload partially some games, so take care and move that folder to another place at first to test.

MMollyMaclachlan 2020-06-16 github

Actually found a fix for this. Under folder "/home/user/.local/share/Steam/" there is two folders with same name but some characters in upper case like that: "SteamApps" and "steamapps".
Deleting the lower case folder "steamapps" solve the problem. Maybe steam could redownload partially some games, so take care and move that folder to another place at first to test.

Be VERY careful when deleting the "steamapps" folder! I've only got one of them, and it's the all lowercase one. For me it contains every appmanifest, the entire "common" folder (which contains almost all of the files for every game I've downloaded), the entire "compatdata" folder (which contains a lot of data relating to Proton, including save files), the entire "shadercache" folder, and the entire "workshop" folder.

AAngryPenguinPL 2020-08-02 github

I can't start The Witcher 3. Interestingly, it happens when I run it from DXVK, but when I set WINE3D (opengl) in launch options, the game starts correctly. After clicking launch in steam. Then steam processing API Vulkan sharders and when this window disappears nothing happens. The witcher.exe process is still running (I can see it in the task manager), unfortunately the game won't start. Worth to add other Proton games works.

In terminal I see error:

wine client error: e8: write: wrong file descriptor

This is on current Proton 5, but when I choose older 4.X then I dont see this error message but Witcher3 still not working (same issue but without error message).

I tried the regular kernel and the one with fsync. Tried launch options noesync and nofsync. Nothing helps.

Here is my Steam OS/hardware info: https://gist.github.com/AngryPenguinPL/1a1b47cd4048a77f824018bc65831cf6

Here is the console log from launch steam to launch witcher: https://gist.github.com/AngryPenguinPL/a68cffcbebfb6ce43b162bd7c2eab4f1

Here is debug log: https://gist.github.com/AngryPenguinPL/5e6477901972472f95dd88f437db02dd

Anyone have any ideas?

NNewbytee 2020-08-02 github

Anyone have any ideas?

sudo dnf install vulkan-loader

Does this help?

AAngryPenguinPL 2020-08-02 github

sudo dnf install vulkan-loader

Does this help?

I have vulkan-loader installed. AlsoI tried on AMDVLK and same error.
Worth to add other Proton games works, I tested The Incredible Adventures of VanHelsingI with OpenGL mode and DirectX11.

4414n 2020-08-02 github

@AngryPenguinPL have you tried clearing the shader cache? I had a similar issue too: https://github.com/ValveSoftware/Proton/issues/607#issuecomment-568767767

AAngryPenguinPL 2020-08-02 github

@AngryPenguinPL have you tried clearing the shader cache? I had a similar issue too: [#607 (comment)](https://github.com/ValveSoftware/Proton/issues/607#issuecomment-568767767)

@414n after remove sharder cache I see steam window - processing API Vulkan but nothing happens. Shaders are not processed, disk usage is nil and so for a long time. I can only click on "Skip".
After skip, in terminal I see again my issue: wine client error: e8: write: wrong file descriptor

but worth to add just before write: wrong file descriptor is this:

>>> Adding process 24044 for game ID 292030
Unable to read VR Path Registry from C:\users\steamuser\Local Settings\Application Data\openvr\openvrpaths.vrpath
Fossilize INFO: Overriding serialization path: "/steam/steamapps/shadercache/292030/fozpipelinesv4/steamapprun_pipeline_cache".
wine client error: e8: write: wrong file descriptor

Maybe this Fossilize INFO is the problem?

VVladislavuz 2020-08-05 github

The game starts from the second or third time.
SteamOS 2.0
steam-292030.log

LLoacoon1 2020-08-12 github

@aeikum I'm pinging you as you told us to do in the surround bug report.
The game runs only in stereo and the sound is low, the voices are even lower.

Aaeikum 2020-08-12 github

@Loacoon1 Thanks. I'll try to get to this sometime, but I'm not sure when.

4414n 2020-12-11 github

Has anyone tried this game with Proton-5.13-4?
I get this error when trying to start it:

25901.439:00c4:00c8:trace:loaddll:build_module Loaded L"C:\\windows\\system32\\xaudio2_7.dll" at 00007F0018080000: builtin
ALSA lib pcm_dmix.c:1108:(snd_pcm_dmix_open) unable to open slave
INFO: OpenAudioDevice failed: ALSA: Couldn't open audio device: Device or resource busy
INFO: Assertion failed: 0 && "Failed to open audio device!"
25901.493:00c4:00c8:trace:seh:dispatch_exception code=c0000005 flags=0 addr=00007F001805D940 ip=1805d940 tid=00c8
25901.493:00c4:00c8:trace:seh:dispatch_exception  info[0]=0000000000000000
25901.493:00c4:00c8:trace:seh:dispatch_exception  info[1]=00000000000000dc
25901.493:00c4:00c8:trace:seh:dispatch_exception  rax=0000000000000000 rbx=0000000002ff0720 rcx=0000000000000000 rdx=000000001805f050
25901.493:00c4:00c8:trace:seh:dispatch_exception  rsi=0000000000000000 rdi=0000000002fdbe90 rbp=0000000000000000 rsp=000000000221e560
25901.493:00c4:00c8:trace:seh:dispatch_exception   r8=000000007d9c7a50  r9=0000000000000001 r10=0000000000000001 r11=000000008262de50
25901.493:00c4:00c8:trace:seh:dispatch_exception  r12=0000000002fdc730 r13=0000000000000000 r14=0000000000000000 r15=0000000002fdc748
25901.493:00c4:00c8:trace:seh:call_vectored_handlers calling handler at 0000000069060770 code=c0000005 flags=0

For some reason, it tries to open an ALSA device and fails to do it.
Here's the complete log: steam-292030.log.gz

LLoacoon1 2020-12-11 github

Apparently it can't find xaudio dll file. You might wanna try installing 'xact' using protontricks in the meantime. Besides, it often fixes my audio issues in games.

MMollyMaclachlan 2020-12-11 github

@414n Worth noting the ALSA issue also affects Hitman 2, as seen here (comment).

4414n 2020-12-12 github

Thanks @Loacoon1, indeed installing xact via winetricks did make the game start.
Now I only ear background voices on my earphones, but that's a minor issue :sweat_smile:

NNewbytee 2020-12-12 github

Has anyone tried this game with Proton-5.13-4?

Strange. Launches just fine for me. Tested both regular Proton 5.13-4 and experimental. I'm on Fedora 32.

Jjzbor 2021-03-29 github

I have a very weird bug where all input keys are shifted one to the left so 'w' is 'e', 's' is 'd' ... Its like this on every Proton Version I tried. I am reinstalling the game right now, lets see if it helps. Has anyone ever encountered this bug?

Aaeikum 2021-03-29 github

@jzbor I haven't heard of that before. Do you have a standard US English keyboard? Some other locale? Anything special about the keyboard?

Jjzbor 2021-03-30 github

@jzbor I haven't heard of that before. Do you have a standard US English keyboard? Some other locale? Anything special about the keyboard?

I have a USB-Keyboard which hardware-wise has a German QWERTZ layout, but software wise I changed it to the standard American layout via setxkbmap...

Kkazzmir 2021-04-29 github

https://gist.github.com/kazzmir/f0fa849ea737fabc9c677b152db0834c system information from steam. I am on ubuntu 20.04 on a laptop with a discrete nvidia card, and when I launch Witcher3 I just get a black screen, and nothing seems to happen. I tried proton experimental, 6.3-2, and 5.13-6 but the same behavior occurs for all of them. Other games like DOOM 2016 do work.

Kkisak-valve maintainer 2021-04-29 github
Kkazzmir 2021-04-29 github

Excellent suggestion! I did this

/usr/share/vulkan/icd.d $ sudo mv intel_icd.x86_64.json intel_icd.x86_64.json.disabled

And the game runs now.

SSakura0134 2021-07-04 github

I am having input delays(Like 1 or more seconds diff) in WItcher 3. But I haven't noticed it in the main menu


Witcher 3 has significant input delay

Issue transferred from https://github.com/ValveSoftware/Proton/issues/4944.
@Sakura0134 posted on 2021-07-05T12:41:48:

Compatibility Report

  • Name of the game with compatibility issues: The Witcher 3
  • Steam AppID of the game: 292030

System Information

I confirm:

  • [x] that I haven't found an existing compatibility report for this game.
  • [x] that I have checked whether there are updates for my system available.

Symptoms

Significant Input delay by 1 to 2 s

Reproduction

Open Witcher 3(Steam Version) and play the game. You will notice input delay by 1 to 2 s

Kkisak-valve maintainer 2021-07-16 github

Witcher 3 weird lighting

Issue transferred from https://github.com/ValveSoftware/Proton/issues/4974.
@ryandbair posted on 2021-07-16T00:56:30:

Compatibility Report

  • Name of the game with compatibility issues: The Witcher 3: Wild Hunt - Game of the Year
  • Steam AppID of the game: 292030

System Information

  • GPU: R9 280 X
  • Driver/LLVM version: Mesa 21.1.4
  • Kernel version: 5.12.15
  • Link to full system information: Gist
  • Proton version: Experimental & 6.3-5

I confirm:

  • [x] that I haven't found an existing compatibility report for this game.
  • [x] that I have checked whether there are updates for my system available.

Symptoms

Lighting in the game is generally a bit weird. Light parts of outdoor scenes are very bright and dark parts are very dark. Caves are near total darkness.

Reproduction

The gamma adjust screen is a good example with Proton Experimental. The background (white) seems to be adjusted but the image in the center is almost always complete black, maybe just flickering into sight here and there.

In game play, just look around. Maybe go into a cave for fun.

steam-292030.log.gz

Kkisak-valve maintainer 2021-07-16 github

Hello @ryandbair, I suspect you're seeing a generation-specific video driver issue. There's been several RADV commits for your video card that have landed between mesa 21.1 and 21.2 and move the driver closer to spec conformance. If it's easy to do, it might be interesting to test how the game behaves with the mesa 21.2 release candidate or a mesa git main build.

Rryandbair 2021-07-16 github

I tested on a build from git this morning (21.3 - 7a1cc56), same issue.

As a correction to my original post, the image in gamma appears only for a split second when the settings are opened. It does not flicker back in after that.

GGoatroth 2021-07-27 github

I'm having the same bug as two others have reported in the past. Neither had any solutions posted to fix the issue. Basically, none of my controls work once I load into the game. Keyboard, mouse, controller, none of them work. I can't even press ESC to bring up the menu. I'm able to use all three to browse through the main menu just fine, and if I minimize the game window to bring up the ESC menu while in the game proper, I'm able to control that menu too. When I look at the keybindings in the options, they all come up as "None" except for quicksave, and the input.settings file is blank save for one entry at the top. If I try to put a fresh input.settings file there and write-protect it, it still shows up in the game as "None" for all keybindings.

I can't generate a Proton log as I'm playing the GOG version with Lutris, but I am using Proton to play the game. The bug does also happen with Lutris's own WINE version. If needed I can paste the Lutris log, but it doesn't seem to have any useful information. Please tell me if I can post any kind of log that would help to figure out this problem, as I'm pretty much unable to play this game at all now until this issue gets resolved.

EDIT: Turns out, it was the Xanmod kernel. I switched to Liquorix (so I could still have fsync support) and the keybindings/no HUD/lack of control issue is completely gone. I suspect something was broken in the latest versions, or maybe it was something in 5.13 that carried over into Xanmod, who knows... all I can say is that it's fixed, for now, and I'm happy.

Kkisak-valve maintainer 2021-11-20 github

The Witcher 3: Wild Hunt

Issue transferred from https://github.com/ValveSoftware/Proton/issues/5331.
@rexhent posted on 2021-11-20T07:25:20:

Whitelist Request

  • Name of the game to be whitelisted: The Witcher 3: Wild Hunt
  • Steam AppID of the game: 292030

System Information

  • GPU: GTX 1050 Mobile
  • Driver/LLVM version: Nvidia 470
  • Distro version: Manjaro Rolling
  • Link to full system information report as Gist:
  • Proton version: Proton Experimental

I confirm:

  • [ Yes] that pressing the Play button in the Steam client is sufficient.
  • [ No] that runtime config options are necessary to run the game.
  • [ No] that no workarounds other than PROTON_* env variables are necessary.

Issues

  • [ Yes] I haven't experienced any issues.
  • [ No] There are no issues left open for this game.
  • [ None] Although I consider the gaming experience equal to Windows there are remaining issues:
AAsciiWolf 2021-11-20 github

Whitelist Request

  • Name of the game to be whitelisted: The Witcher 3: Wild Hunt
  • Steam AppID of the game: 292030

System Information

I confirm:

  • [x] that pressing the Play button in the Steam client is sufficient.
  • [ ] that runtime config options are necessary to run the game.
  • [ ] that no workarounds other than PROTON_* env variables are necessary.

Issues

  • [x] I haven't experienced any issues.
  • [ ] There are no issues left open for this game.
  • [ ] Although I consider the gaming experience equal to Windows there are remaining issues:

Everything works fine out-of-box just like it was a native game.

Kkisak-valve maintainer 2021-11-29 github

The Witcher 3 freezes at the start of game

Issue transferred from https://github.com/ValveSoftware/Proton/issues/5363.
@x-crowbar-x posted on 2021-11-29T13:18:39:

Compatibility Report

  • Name of the game with compatibility issues: The Witcher 3: Wild Hunt
  • Steam AppID of the game: 292030

System Information

  • CPU: AMD Ryzen 3 4300U with Radeon Graphics
  • GPU: AMD RENOIR (integrated) 1 GiB.
  • Driver: AMD AMD RENOIR (DRM 3.42.0, 5.15.5-arch1-1, LLVM 13.0.0)
  • Driver Version: 4.6 (Compatibility Profile) Mesa 21.2.5
  • OpenGL Version: 4.6
  • Kernel version: 5.15.5-arch1-1
  • Link to full system information report as Gist
  • Proton version: 6.3-8

I confirm:

  • [🗸] that I haven't found an existing compatibility report for this game.
  • [🗸] that I have checked whether there are updates for my system available.

steam-292030.log

Symptoms

I started the game and everything was fine until I started a new game. I skipped the videos and it freezed after a couple of seconds in gameplay. It always freezes at the same spot. I ran it in the lowest settings and even chose the lowest resolution.
Also, the laptop does not even get hotter and the fan is almost quiet, as if it doesn't try to load anything, but I can hear the sounds perfectly.

P.S. I found videos of people playing the game in 30-34 fps on my hardware on Windows, so it should also work on Linux.

Reproduction

Open the game and click on "NEW GAME", skip the video, or don't, it does not matter.

Kkisak-valve maintainer 2021-11-29 github

Hello @x-crowbar-x, looking at your Proton log, err: DxvkSubmissionQueue: Command submission failed: VK_ERROR_DEVICE_LOST looks like the line of interest, and your system information tells us that your system is configured to use AMDVLK. Can you test if game has the same issue when run with mesa/RADV?

Xx-crowbar-x 2021-11-29 github

@kisak-valve Well, I will try, but I do not know how to run it with mesa/RADV, to be honest. Can you tell me how?

Kkisak-valve maintainer 2021-11-29 github

For your distro, I think you can get mesa/RADV by installing the vulkan-radeon and lib32-vulkan-radeon system packages. After that, removing the amdvlk and lib32-amdvlk system packages or setting the game's launch options to AMD_VULKAN_ICD=RADV %command% should get the game to use mesa/RADV. (https://wiki.archlinux.org/title/AMDGPU https://wiki.archlinux.org/title/Vulkan#Selecting_Vulkan_driver)

Xx-crowbar-x 2021-11-29 github

Thank you very much @kisak-valve! It works just fine now!

Kkisak-valve maintainer 2021-12-17 github

The Witcher 3: Wild Hunt (292030)

Issue transferred from https://github.com/ValveSoftware/Proton/issues/5417.
@minionprocyk posted on 2021-12-17T19:50:40:

Compatibility Report

  • Name of the game with compatibility issues: The Witcher 3
  • Steam AppID of the game: 292030

System Information

  • GPU: AMD 5700XT
  • Driver/LLVM version: 21.3.1
  • Kernel version: 5.15.6-200.fc35.x86_64
  • Link to full system information report as Gist:
  • Proton version: proton-6.3-8

I confirm:

  • [x] that I haven't found an existing compatibility report for this game.
  • [x] that I have checked whether there are updates for my system available.

steam-292030.log

Symptoms

Game crashes on launch. This started happening after upgrading my CPU/Motherboard.

Reproduction

Launch Game

IIlluminat10n 2021-12-23 github

Today I couldn't run the game. It used to run fine just yesterday. I don't think I even updated anything.. The error is as follows:

ERROR: ld.so: object '/home/lumi/.local/share/Steam/ubuntu12_32/gameoverlayrenderer.so' from LD_PRELOAD cannot be preloaded (wrong ELF class: ELFCLASS32): ignored.
ERROR: ld.so: object '/home/lumi/.local/share/Steam/ubuntu12_32/gameoverlayrenderer.so' from LD_PRELOAD cannot be preloaded (wrong ELF class: ELFCLASS32): ignored.
GLib-CRITICAL **: 01:40:40.890: unquote_string_inplace: assertion 'err == NULL || *err == NULL' failed
GameAction [AppID 292030, ActionID 1] : LaunchApp changed task to ProcessingInstallScript with ""
pressure-vessel-adverb[27718]: W: Unable to generate locales: Child process exited with code 1
Installing breakpad exception handler for appid(steam)/version(1639697812)
Installing breakpad exception handler for appid(steam)/version(1639697812)

I tried everything. Reinstalled drivers, steam, the game, packages...
I have an optimus laptop with ryzen 5 4600H and GTX 1650 Ti running manjaro linux.
Gist: https://gist.github.com/I-dont-need-name/4fabdef5063dbdcfd4df4f0cbef3ca03

Ttriffid 2021-12-23 github

With a bit of googling for "GLib-CRITICAL **: 01:40:40.890:
unquote_string_inplace: assertion 'err == NULL || *err == NULL' failed",
https://github.com/ValveSoftware/steam-for-linux/issues/7664 suggests it's
an issue with desktop shortcuts not making steam update proton libs or
something, not specific to Witcher 3 nor Proton's compatibility layer, but
some quirk of steam runtime management.

However, there's also a good chance that your error occurred earlier in the
log, and by showing just the tail you'll end up on a wild goose chase - why
not add the full log to your gist?

On Thu, 23 Dec 2021 at 15:21, I-dont-need-name @.***>
wrote:

Today I couldn't run the game. It used to run fine just yesterday. I don't
think I even updated anything.. The error is as follows:

ERROR: ld.so: object '/home/lumi/.local/share/Steam/ubuntu12_32/gameoverlayrenderer.so' from LD_PRELOAD cannot be preloaded (wrong ELF class: ELFCLASS32): ignored.
ERROR: ld.so: object '/home/lumi/.local/share/Steam/ubuntu12_32/gameoverlayrenderer.so' from LD_PRELOAD cannot be preloaded (wrong ELF class: ELFCLASS32): ignored.
GLib-CRITICAL **: 01:40:40.890: unquote_string_inplace: assertion 'err == NULL || *err == NULL' failed
GameAction [AppID 292030, ActionID 1] : LaunchApp changed task to ProcessingInstallScript with ""
pressure-vessel-adverb[27718]: W: Unable to generate locales: Child process exited with code 1
Installing breakpad exception handler for appid(steam)/version(1639697812)
Installing breakpad exception handler for appid(steam)/version(1639697812)

I tried everything. Reinstalled drivers, steam, the game, packages...
I have an optimus laptop with ryzen 5 4600H and GTX 1650 Ti running
manjaro linux.
Gist:
https://gist.github.com/I-dont-need-name/4fabdef5063dbdcfd4df4f0cbef3ca03

—
Reply to this email directly, view it on GitHub
https://github.com/ValveSoftware/Proton/issues/607#issuecomment-1000097689,
or unsubscribe
https://github.com/notifications/unsubscribe-auth/AABICGXR4FRZUGVSVXYGPI3USLEXJANCNFSM4FRSLUZQ
.
Triage notifications on the go with GitHub Mobile for iOS
https://apps.apple.com/app/apple-store/id1477376905?ct=notification-email&mt=8&pt=524675
or Android
https://play.google.com/store/apps/details?id=com.github.android&referrer=utm_campaign%3Dnotification-email%26utm_medium%3Demail%26utm_source%3Dgithub.

You are receiving this because you were mentioned.Message ID:
@.***>

IIlluminat10n 2021-12-23 github

why not add the full log to your gist?

Sorry, here is the log from running steam in console until I launch the game and nothing happens (it says "running" but nothing is running):

 $ steam                                                                                                                                                                         ✔ 
steam.sh[16713]: Running Steam on manjarolinux 21.2.0 64-bit
steam.sh[16713]: STEAM_RUNTIME is enabled automatically
setup.sh[16786]: Steam runtime environment up-to-date!
steam.sh[16713]: Steam client's requirements are satisfied
[2021-12-23 09:54:44] Startup - updater built Dec 16 2021 22:39:04
Installing breakpad exception handler for appid(steam)/version(1639697812)
[2021-12-23 09:54:44] Loading cached metrics from disk (/home/lumi/.local/share/Steam/package/steam_client_metrics.bin)
[2021-12-23 09:54:44] Using the following download hosts for Public, Realm steamglobal
[2021-12-23 09:54:44] 1. https://cdn.cloudflare.steamstatic.com, /client/, Realm 'steamglobal', weight was 100, source = 'update_hosts_cached.vdf'
[2021-12-23 09:54:44] 2. https://cdn.akamai.steamstatic.com, /client/, Realm 'steamglobal', weight was 100, source = 'update_hosts_cached.vdf'
[2021-12-23 09:54:44] 3. http://media.steampowered.com, /client/, Realm 'steamglobal', weight was 1, source = 'baked in'
Installing breakpad exception handler for appid(steam)/version(1639697812)
[2021-12-23 09:54:44] Verifying installation...
[2021-12-23 09:54:44] Verification complete
Loaded SDL version 2.0.19-6963453
/home/lumi/.themes/Materia-Manjaro-Dark-Solid-gtk/gtk-2.0/widgets/entry:70: error: unexpected identifier 'direction', expected character '}'

(steam:16870): Gtk-WARNING **: 09:54:44.790: Unable to locate theme engine in module_path: "adwaita",
/home/lumi/.themes/Materia-Manjaro-Dark-Solid-gtk/gtk-2.0/widgets/styles:36: error: invalid string constant "combobox_entry", expected valid string constant
Installing breakpad exception handler for appid(steam)/version(1639697812)
steamwebhelper.sh[16923]: Runtime for steamwebhelper: defaulting to /home/lumi/.local/share/Steam/ubuntu12_64/steam-runtime-heavy
Installing breakpad exception handler for appid(steam)/version(1639697812)
Installing breakpad exception handler for appid(steam)/version(1639697812)
Installing breakpad exception handler for appid(steam)/version(1639697812)
Installing breakpad exception handler for appid(steam)/version(1639697812)
CAppInfoCacheReadFromDiskThread took 20 milliseconds to initialize
Installing breakpad exception handler for appid(steam)/version(1639697812)
Installing breakpad exception handler for appid(steam)/version(1639697812)
Installing breakpad exception handler for appid(steam)/version(1639697812)
Installing breakpad exception handler for appid(steam)/version(1639697812)
Installing breakpad exception handler for appid(steam)/version(1639697812)
Installing breakpad exception handler for appid(steam)/version(1639697812)
Installing breakpad exception handler for appid(steam)/version(1639697812)
Installing breakpad exception handler for appid(steam)/version(1639697812)
Installing breakpad exception handler for appid(steam)/version(1639697812)
Installing breakpad exception handler for appid(steam)/version(1639697812)
Installing breakpad exception handler for appid(steam)/version(1639697812)
Proceed to auto login
Installing breakpad exception handler for appid(steam)/version(1639697812)
Installing breakpad exception handler for appid(steam)/version(1639697812)

(steam:16870): GLib-CRITICAL **: 09:54:45.435: g_main_context_push_thread_default: assertion 'acquired_context' failed

(steam:16870): GLib-CRITICAL **: 09:54:45.436: g_main_context_pop_thread_default: assertion 'g_queue_peek_head (stack) == context' failed
Opted-in Controller Mask for AppId 0: 0
Installing breakpad exception handler for appid(steam)/version(1639697812)

(steam:16870): Gtk-WARNING **: 09:54:45.737: gtk_disable_setlocale() must be called before gtk_init()
Installing breakpad exception handler for appid(steam)/version(1639697812)
Installing breakpad exception handler for appid(steam)/version(1639697812)
roaming config store loaded successfully - 235 bytes.
migrating temporary roaming config store
BRefreshApplicationsInLibrary 1: 9ms
Installing breakpad exception handler for appid(steam)/version(1639697812)
Failed to init SteamVR because it isn't installed
ExecCommandLine: "'/home/lumi/.local/share/Steam/ubuntu12_32/steam'"
System startup time: 3.82 seconds
BuildCompleteAppOverviewChange: 120
RegisterForAppOverview 1: 12ms
RegisterForAppOverview 2: 13ms
GameAction [AppID 292030, ActionID 1] : LaunchApp changed task to ProcessingInstallScript with ""
pressure-vessel-adverb[17465]: W: Unable to generate locales: Child process exited with code 1
GameAction [AppID 292030, ActionID 1] : LaunchApp changed task to SynchronizingCloud with ""
GameAction [AppID 292030, ActionID 1] : LaunchApp changed task to SiteLicenseSeatCheckout with ""
GameAction [AppID 292030, ActionID 1] : LaunchApp changed task to CreatingProcess with ""
GameAction [AppID 292030, ActionID 1] : LaunchApp waiting for user response to CreatingProcess ""
GameAction [AppID 292030, ActionID 1] : LaunchApp continues with user response "CreatingProcess"
Opted-in Controller Mask for AppId 292030: 0
/bin/sh\0-c\0/home/lumi/.local/share/Steam/ubuntu12_32/reaper SteamLaunch AppId=292030 -- '/home/lumi/.local/share/Steam/steamapps/common/SteamLinuxRuntime_soldier'/_v2-entry-point --verb=waitforexitandrun -- '/home/lumi/.local/share/Steam/steamapps/common/Proton - Experimental'/proton waitforexitandrun  '/home/lumi/.local/share/Steam/steamapps/common/The Witcher 3/bin/x64/witcher3.exe'\0
Game process added : AppID 292030 "/home/lumi/.local/share/Steam/ubuntu12_32/reaper SteamLaunch AppId=292030 -- '/home/lumi/.local/share/Steam/steamapps/common/SteamLinuxRuntime_soldier'/_v2-entry-point --verb=waitforexitandrun -- '/home/lumi/.local/share/Steam/steamapps/common/Proton - Experimental'/proton waitforexitandrun  '/home/lumi/.local/share/Steam/steamapps/common/The Witcher 3/bin/x64/witcher3.exe'", ProcID 17474, IP 0.0.0.0:0
chdir /home/lumi/.local/share/Steam/steamapps/common/The Witcher 3/bin/x64
ERROR: ld.so: object '/home/lumi/.local/share/Steam/ubuntu12_32/gameoverlayrenderer.so' from LD_PRELOAD cannot be preloaded (wrong ELF class: ELFCLASS32): ignored.
ERROR: ld.so: object '/home/lumi/.local/share/Steam/ubuntu12_64/gameoverlayrenderer.so' from LD_PRELOAD cannot be preloaded (wrong ELF class: ELFCLASS64): ignored.
GameAction [AppID 292030, ActionID 1] : LaunchApp changed task to WaitingGameWindow with ""
ERROR: ld.so: object '/home/lumi/.local/share/Steam/ubuntu12_32/gameoverlayrenderer.so' from LD_PRELOAD cannot be preloaded (wrong ELF class: ELFCLASS32): ignored.
ERROR: ld.so: object '/home/lumi/.local/share/Steam/ubuntu12_32/gameoverlayrenderer.so' from LD_PRELOAD cannot be preloaded (wrong ELF class: ELFCLASS32): ignored.
ERROR: ld.so: object '/home/lumi/.local/share/Steam/ubuntu12_32/gameoverlayrenderer.so' from LD_PRELOAD cannot be preloaded (wrong ELF class: ELFCLASS32): ignored.
GameAction [AppID 292030, ActionID 1] : LaunchApp changed task to Completed with ""
pid 17478 != 17475, skipping destruction (fork without exec?)
pressure-vessel-adverb[17647]: W: Unable to generate locales: Child process exited with code 1
Installing breakpad exception handler for appid(steam)/version(1639697812)
Installing breakpad exception handler for appid(steam)/version(1639697812)

The error also looks a bit different this time

UPD: it seems to be a vulkan-related problem. The game runs fine if I set optimus-manager to nvidia mode. If I run it in integrated mode (what I used to and want to do) it doesn't work and gives this error. However, it runs on internal gpu with VK_ICD_FILENAMES=/usr/share/vulkan/icd.d/radeon_icd.i686.json:/usr/share/vulkan/icd.d/radeon_icd.x86_64.json steam.

UPD2: everything works as expected with proton 5.0. Better than nothing but still kinda sucks to use outdated version to have things working.

Ppoperigby 2022-01-16 github

The game takes a really long time to load for me, almost 30 seconds. I'm on an SSD though. I'm using mods, but I tried uninstalling them and it still happened. On a computer running Windows, but with a old slow HDD, it was quite fast.

System info

Sshmerl 2022-01-16 github

I think the game has some issues with too many saves. Try pruning that.

Ppoperigby 2022-01-16 github

Not sure if that's it. I deleted most of my saves.

Nntorresalberto 2022-01-27 github

Hello, the game is crashing on launch (running on fedora 35 amd cpu + amd dgpu rx6800m).
It sends me back to the login screen (of the display manager).

I've attached the proton log: steam-292030.log

I can get tt works if I use PROTON_USE_WINED3D=1 on the integrated graphics but it won't work on the discrete card (by settingsDRI_PRIME=1).

edit: solved, see comment below.

Mminionprocyk 2022-01-28 github

hi @ntorresalberto I was experiencing a crash on launch, presumably because of missing libraries after updating my fedora install. I resolved this by installing the flatpak version of steam which has all required dependencies bundled together. This fixed CS:GO and The Witcher 3 for me. Hope this works for you, cheers.

Nntorresalberto 2022-01-28 github

hi @ntorresalberto I was experiencing a crash on launch, presumably because of missing libraries after updating my fedora install. I resolved this by installing the flatpak version of steam which has all required dependencies bundled together. This fixed CS:GO and The Witcher 3 for me. Hope this works for you, cheers.

I'll give it a try, but this is the only game crashing on launch. Everything else seems to be working fine.

edit: I tried the flatpak and I get the same error. But I've since discovered there are other games affected by (what it seems to me) the same problem, so I guess it is not exclusive to the witcher 3.

edit2: just figured out for some reason Witcher 3 requires a compositor (I was running it on i3 windows manager without a compositor). It worked fine on gnome (everything else seemed to work fine, that's why I didn't think about it before).
So a simple fix (in i3) is launching a simple compositor like picom (successor of compton) and it works fine. Don't know if this is related to having freesync enabled (vsync off) in X11.

Kkisak-valve maintainer 2022-04-21 github

Hello @ryandbair, can you retest if you're still seeing the lighting issue you've mentioned at https://github.com/ValveSoftware/Proton/issues/607#issuecomment-881103994 with a current point release of mesa or mesa git main?

Rryandbair 2022-04-24 github

@kisak-valve I'll try to get around to that soon. I believe this is more of a drm/color space issue and I haven't seen any movement on that front.

The core of the issue seems to be that my display is getting sent a full RGB signal but is not expecting it resulting in colors getting crushed. If I take an in-game screen shot and transfer it to my desktop with Intel graphics it looks fine.

In Windows, it defaults to broadcast RGB which looks fine but if I manually select full RGB I see the same problem.

Kkisak-valve maintainer 2022-05-27 github

Tracking note: Mesa/RADV labels tentatively dropped. @ryandbair, if your issue is indeed on the drm/kernel side, there's a better chance there will be progress if you report the issue you're seeing directly to your video driver vendor.

PPatola 2022-12-14 github

Something very odd is happening with the proton issue search, I try the search with "Witcher 3 292030" and always get almost 50 pages of issues, instead of the game issue. That might be the reason there are no new updates on this ticket since May 27. Tried other search combinations and never got to the ticket, had to resort to search by using google and the "site:" modifier.

I'm running the game with two mods from Nexus Mods: modNoTimeForGwent and modOver9000. I was also using modUnlimitedDurability but this one would not compile after this new 50 GB update, so I took it off for the game to run.

Both in the Steam Deck and my gaming PC with RX 6800 XT, it renders the new launcher without problems. It can log in successfully to the GOG account and start the game in DirectX 12 and even get to the menu screen, where Geralt is meditating. However when I start the game (by clicking "continue", since I have a savegame) it starts loading and after a while crashes to the desktop. Same thing both in the Deck and the PC with arch. Attached the log from the PC and the Deck in DirectX 12 mode.

steam-292030-linux-pc.log
steam-292030-steamdeck.log

So far, it seems to run ok in DirectX 11 mode (both on the PC and the Deck), but I haven't tested for long, some people are saying it also crashes after a while.

Haven't even tried activating RayTracing yet, and if I do, I'll do it first on my RTX 3070 laptop. My results with Portal with RTX were not good on AMD.

Kkisak-valve maintainer 2022-12-14 github

Something very odd is happening with the proton issue search

The Witcher 3 is used as an example in hidden text of the compatibility report template. A Github search for 292030 gave this compatibility report as the first hit.

PPatola 2022-12-14 github

Something very odd is happening with the proton issue search

The Witcher 3 is used as an example in hidden text of the compatibility report template. A Github search for 292030 gave this compatibility report as the first hit.

For some reason, not here:
image

Kkisak-valve maintainer 2022-12-14 github

Issue report filtering is apparently subtly different from the site's search function: https://github.com/ValveSoftware/Proton/search?q=292030&type=issues

Sshmerl 2022-12-14 github

Just tried version 4.0 with recent vkd3d-proton and Wine 8.0~rc1 (I'm launching dx12 variant of witcher3.exe).

The game crashes with such error after initial cutscene in Kaer Morhen playing alright:

Assertion failed: !status, file dlls/winevulkan/loader_thunks.c, line 3145

Just for the reference, it's from here.

NNiedzwiedzw 2022-12-14 github

The game stopped working after the next-gen patch sadly

PPatola 2022-12-14 github

The game stopped working after the next-gen patch sadly

How so? What is your configuration (gist), do you use any mods? What happens when you press play, are you able to see the CD Projekt Red Launcher? Which proton are you using? Do you have some dialog error? Have tried with PROTON_LOG=1 %command% as launch parameter and got the steam-292030.log file?

NNiedzwiedzw 2022-12-14 github

@Patola
After changing to DX11 (recommended for older computers) it launched. The game shows 60FPS, but it stutters like hell. No mods.

Ooakad 2022-12-14 github

I have a Linux machine here which used to run Witcher 3 just fine (main game, both DLCs, whatever else) as recently as couple of months ago. No changes to the setup, apart from the usual updates.

With the present update:

  1. Proton Experimental: red-launcher crashes in verify phase, game executable sort of starts and sits there with a blank, black window
  2. Proton 7.0-5: same as Proton Experimental
  3. Proton 6.3-8/dx12: red-launcher starts and works as expected, shows all the offers and such. Game crashes hard, unable to show a proper crash report (the crash report also crashes)
  4. Proton 6.3-8/dx11: game actually works (not too smoothly). :-)

red-launcher crash looks as following:

2022-12-14 20:29:39.390 [Information][ (0)] [TID 308][red-prelauncher]: Starting
REDlauncher verification. Program: 'C:/users/steamuser/AppData/Local/Programs/C
D Projekt Red/REDlauncher/REDlauncher.exe'. Arguments: '--launcher-verify'.
2022-12-14 20:29:39.675 [Error][ (0)] [TID 308][red-prelauncher]: REDlauncher ve
rification failed. Falling back to launching the game. Error: Error Message: RED
launcher crashed. Process error: Process crashed (code: 1)
Stack Trace
7: UNK - 0x4627d7
6: UNK - 0x40fca5
5: UNK - 0x4116b9
4: UNK - 0x4b9287
3: BaseThreadInitThunk - 0x7b62ddb0
2: RtlUserThreadStart - 0x7bc5d3d7
1: RtlCreateUserThread - 0x7bc5db80
0: UNK - 0x0
-----
.

Jjorgicio 2022-12-14 github

Just got the Next-Gen update, the launcher works, and the first try it crashes at start (no matter the Directx version). I had a mod installed, so that was the culprit. So if you have mods, you may remove them or the game will crash at start.

After that, the game launches, but only Directx11 version works. Directx12 version starts, but there's a black screen only. Also, Directx11 should come with FSR too (at least, 1.0). The issue happened with any Proton I tried (GE, TKG, Experimental, Stable).

Anyways, here's the log.

steam-292030.log

AAngryPenguinPL 2022-12-14 github

For me in next gen update launcher still crashing.
I even reverted update and in beta tab pick old game (classic, patch 1.32) and it also crashing at launch due this new launcher...
But classic version works, if I add to steam just a witcher.exe as a non-steam game...

Mmatyat 2022-12-14 github

Seems to crash in dx11 and dx12 mode if Nvidia hairworks is turned on

Mmbriar 2022-12-14 github

DX12 mode should work on latest Experimental Bleeding Edge now. Hairworks apparently needs d3dcompiler_47.

Sshmerl 2022-12-14 github

I can confirm that latest vkd3d-proton master fixes the issue. Performance of the game is surprisingly bad though in this version.

Zzaps166 2022-12-14 github

Hairworks apparently needs d3dcompiler_47

This fixes the issue with HairWorks. However:

  • NVIDIA + DX12 + HairWorks = NVRM: Xid (PCI:0000:01:00): 69, pid=25360, name=witcher3.exe, Class Error: ChId 0016, Class 0000c597, Offset 00000274, Data 00005de5, ErrorCode 000000c1.
  • Works properly on RADV + DX12 + HairWorks.
  • Works properly on NVIDIA/RADV + DX11 + HairWorks.
Ssim590 2022-12-14 github

In case anyone goes into the same issue as I did with DirectX12 displaying blackscreen and if you want to avoid using the red launcher to configure DirectX11 as a pre step for starting the game, see this post.

BBlisto91 2022-12-14 github

Note with the d3d11 version there might be a d3dcompiler issue that causes invisible Geralt hair and performance problems.
Dunno if it's only when hairworks is turned on. Can be worked around by installing Windows version of d3dcompiler_47 through proton-/winetricks.

TTea23 2022-12-15 github

Working okay in proton-next on a 6700XT with Hairworks enabled. Raytracing options appear to be disabled, however. I also had a hard crash when Geralt rushes the Gryphon at the start of the game.

Sshmerl 2022-12-15 github

Not sure if ray tracing in radv will work for for it, but to try it you should set these:

RADV_PERFTEST='rt,gpl'
VKD3D_CONFIG='dxr11'
TTea23 2022-12-15 github

@shmerl Thanks! That did indeed get RT enabled, but led to a full desktop lock up. Reading into it a bit more I see it's entirely experimental at the driver level, not something I knew about (I'm completely new to RT-capable GPUs).

Sshmerl 2022-12-15 github

@Tea23: You can report the crash to Mesa with some details so developers could note the issue.

Sshmerl 2022-12-15 github

Observation on performance.

In DX12 mode the game doesn't saturate the GPU fully (even on lower settings) so performance is really bad (and that's even with high end CPU). It's a problem with the game itself I think, not with vkd3d-proton. In DX11 mode the game flies loading the GPU completely even on Ultra+ settings, so performance is great.

I recommend avoiding DX12 mode for this update and play it in DX11 one, until they fix the performance, if that ever happens.

Apparently as some reviewers pointed out, DX12 mode is not a real engine rewrite, they use some kind of DX11 over DX12 layer there, which in itself doesn't perform well (add on top of that the need to translate DX12 into Vulkan, so you get the idea why performance is tanked on Linux even more).

Mmbriar 2022-12-15 github

It looks like the game ships d3d11on12, so it probably uses it at least for some parts of the rendering when running on d3d12, might explain the bad performance.

BBlisto91 2022-12-15 github

According to dxvk/vkd3d-proton devs it does not appear upon first inspection to use d3d11on12 and that it's likely just a leftover thats been forgotten to be removed. (Haven't checked myself)

Apparently its cpu bound because

it's cpu-bound because they just derped all their d3d12 code into one thread

Sshmerl 2022-12-15 github

@Blisto91: Where is that comment from?

BBlisto91 2022-12-15 github

Linux Gaming Dev discord

Ddsalt 2022-12-15 github

@shmerl Thanks! That did indeed get RT enabled, but led to a full desktop lock up. Reading into it a bit more I see it's entirely experimental at the driver level, not something I knew about (I'm completely new to RT-capable GPUs).

That crash, or a very similar one, has already been reported – I recommend comparing details with that and adding info there if needed.

AAngryPenguinPL 2022-12-15 github

Anyone know how to resolve launcher crashing in Classic edition? (available from beta tab, classic = last patch before 4.0)

LLoacoon1 2022-12-15 github

Is anyone else having a broken graphical settings page since the update?
The settings are invisible, all I can see are the global presets.

Ppoperigby 2022-12-15 github

The game won't launch for me regardless of DirectX version. It will get to the launcher, but clicking play just causes the process to crash.

I deleted everything in the game folder and reinstalled, I removed my settings from the documents folder in the prefix, and I tried both Proton Experimental and GE-Proton7-42.

OS: ArchLinux
GPU: RX 5700XT
Mesa version: 22.2.3

TTea23 2022-12-15 github

@poperigby Try installing Proton Hotfix, it seems there is a specific W3 fix in there right now.

Ppoperigby 2022-12-15 github

@poperigby Try installing Proton Hotfix, it seems there is a specific W3 fix in there right now.

That doesn't seem to work, unfortunately.

Ppoperigby 2022-12-15 github

steam-292030.log

Here's the Proton log.

TTea23 2022-12-15 github

steam-292030.log

Here's the Proton log.

In that log, you are still running Proton Experimenal. You'll need to switch to Proton Hotfix in Witcher 3's compatibility settings.

Ppoperigby 2022-12-15 github

Right, I did run it before with the hotfix and it still didn't work, but I'll post the Proton hotfix log too.

steam-292030.log

Kkisak-valve maintainer 2022-12-15 github

Hello @poperigby, err:module:import_dll Library DINPUT8.dll (which is needed by L"Z:\\home\\cassidy\\.local\\share\\Steam\\steamapps\\common\\The Witcher 3\\bin\\x64\\witcher3.exe") not found looks like the line of interest from your log.

Maybe backup your save files and reset the wineprefix?

Ppoperigby 2022-12-15 github

That fixed it! Thank you.

TTea23 2022-12-18 github

With the patches in proton-hotfix I'm having a pretty decent time playing this, except for one while crash which I can't quite pin down. I attached the proton log of my last session in case it helps (I've had to compress it though because it's 67MB and github won't allow me to upload >25). steam-292030.zip

The crash appears randomly, so far it has been in cutscenes only. The frame will freeze but music and some sound effects will continue as normal. I become unable to exit the game, even if I SSH into my machine and kill -9 it, so I have to hard shutdown.

LLoacoon1 2022-12-21 github

I still can't see the graphical settings. Every other menu works but the graphical settings stay transparent. Any idea?

Hhsmalley 2022-12-24 github

I'm getting a problem where the game crashes and turns into a zombie that I have to do a kill -9 to exit. I've tried everything I can think of from deleting the shader cache and proton hotfix, next, experimental, various GE builds, etc. I deleted the prefix each time I switched proton versions too just to be extra sure. I even reinstalled just to make sure. So I'm at a complete loss on why the nextgen update just doesn't work.

steam-292030.log

Hhsmalley 2022-12-24 github

Just on a whim I tried doing PROTON_SET_GAME_DRIVE=1 and it works.... Why? I have no idea.... Here's the log of it working in a fresh prefix with default proton 7.0.5.
steam-292030.log

I did have some instability while playing it so it I did switch to Proton-GE7-43 and everything is smooth.

TTJonCooper 2022-12-25 github

Ubuntu 22.04.1 LTS
NVIDIA RTX 3060Ti
NVIDIA Drivers: 525.60.11
Launch Options: PROTON_ENABLE_NVAPI=1 PROTON_HIDE_NVIDIA_GPU=0 VKD3D_CONFIG=dxr %command%

If ray-tracing is turned on, all of the menus work but the second you try to launch the game (Continue/New Game) - it freezes and has to be force quit. It never actually loads any of the game world. Sometimes the game reboots back to the start menu/splash screens.

This is happening on Proton Experimental, Proton Hotfix, and the latest ProtonGE (7-43).

Ray-tracing off, all other settings maxed out - no issues, high FPS.

Any ideas?

AAsciiWolf 2022-12-30 github

Anyone know how to resolve launcher crashing in Classic edition? (available from beta tab, classic = last patch before 4.0)

Got it working! You just have to select "DirectX 11" version in the red launcher. :-)

The game version sadly seems to always default to DX12 when using --launcher-skip.

BBlisto91 2022-12-30 github

You can change the default DirectX when launcher doesn't appear in a file afaik
https://steamcommunity.com/app/292030/discussions/0/6126615404779495660/

Ttripzero 2023-01-11 github

Game starts with a blank black window for me with DX12, proton experimental and latest mesa 23.0-devel and Intel Arc A750. DX11 seems to work though. I have an issue in mesa for it here: https://gitlab.freedesktop.org/mesa/mesa/-/issues/8057

Kkisak-valve maintainer 2023-01-17 github

The Witcher 3 (v4) + Radeon + Mesa = Kernel panic

Issue transferred from https://github.com/ValveSoftware/Proton/issues/6461.
@xanscale posted on 2023-01-17T18:38:11:

Compatibility Report

  • Name of the game with compatibility issues: The witcher 3 v4.00
  • Steam AppID of the game: 292030

System Information

  • GPU: AMD AMD Radeon RX 6600 (navi23, LLVM 15.0.5, DRM 3.49, 6.1.6-060106-generic)
  • Driver/LLVM version: 4.6 (Compatibility Profile) Mesa 23.1.0-devel (git-bb003d4 2023-01-17 jammy-oibaf-ppa)
  • Kernel version: 6.1.6-060106-generic
  • Link to full system information report: system-info.txt
  • Proton version: Latest Experimental

I confirm:

  • [X] that I haven't found an existing compatibility report for this game.
  • [X] that I have checked whether there are updates for my system available.

steam log

steam-292030.log

Symptoms / Reproduction

With this configuration running the game in DX12 mode after few minutes of perfect gaming, we receive a kernel panic (or just machine become totally frozen)
this is the dmesg log crash.log
i noticed this bug with standard Ubuntu 20.04 packages, i updated kernel and mesa with last available and nothing changed

Also on https://www.protondb.com/app/292030 there is some other user with same config with same problem

Ddanisztls 2023-01-28 github

Arch Linux, RX 6600 XT, Mesa Drivers, Proton Experimental

System Info
Proton Log

I'm able to load a game and play with DX12 but after a moment the X server crashes leaving the GPU in a broken state (requiring a full reboot).

Xxanscale 2023-01-31 github

@kisak-valve the problem here https://github.com/ValveSoftware/Proton/issues/607#issuecomment-1385870801
the problem involves a kernel panic and requires a full reboot of the machine. I think it doesn't concern proton directly but more like the AMDgpu kernel module or the Mesa drivers. should be more evidence, the other issues in this thread are either dated or otherwise less relevant

https://gitlab.freedesktop.org/drm/amd/-/issues/2300

CChaosBlades 2023-02-16 github

With the latest patch v4.01 DirectX 12 mode gives me almost exactly half the performance of DirectX 11 mode at the same graphics settings. Everything extra that DX12 provides is disabled. Everything else is maxed out. Does not matter if I lower everything to the lowest setting. It still gives me almost exactly half performance. This is on a 6900 XT. Still have the same issue with the latest Hot fix that released yesterday that supposedly fixed performance issues. In both modes I will see around 99% GPU usage. Before the patch, both directX modes had roughly the same performance. DirectX 11 mode still gives me the same performance as before v4.01

Sshmerl 2023-02-16 github

I think it's a game issue. Hotfix didn't fix anything. It's a good thing you are seeing 99% GPU usage. I see only around 80% in DX12 mode, that's why performance there is bad.

Zzaps166 2023-05-01 github

Regressions with Proton 8.0-1 (Game v4.02, DX11):

  • after alt-tab the game runs in windowed mode all the time (taskbar is visible). Changing the game settings (windowed/borderless/fullscreen) doesn't help. Also using WM's hotkey to go fullscreen doesn't help (looks like window size is 1px less than screen size). Only game restart helps. No problems with Proton 7.0-6.
  • DualSense PS5 gamepad is not detected by the game when it's connected while game is already running. After reconnection with Proton 7.0-6 gamepad was working with vibrations, but no haptics.
Kkisak-valve maintainer 2023-05-01 github

Hello @zaps166, please add PROTON_LOG=1 %command% to the game's launch options, reproduce the regressions, and attach the generated $HOME/steam-$APPID.log to this issue report as a file. (Proton logs compress well if needed.)

Also, please copy your system information from Steam (Steam -> Help -> System Information) and put it in a gist, then include a link to the gist in this issue report. If you're using the Steam client beta, then you may need to temporarily switch to the non-beta client to do that easily.

Zzaps166 2023-05-01 github
Kkisak-valve maintainer 2023-05-01 github

This looks like a line of interest from the Proton log: err:plugplay:handle_bus_relations Failed to enumerate child devices, status 0xc00000bb. this is immediately followed by an access violation (c0000005) and wine: Unhandled page fault on read access to FFFFFFFFFFFFFFFF at address 000000031CAA2F46 (thread 0348), starting debugger....

Zzaps166 2023-05-01 github

Also I see this x11drv:update_net_wm_fullscreen_monitors hwnd 0x2008e at (-3,-22)-(3843,2162) is outside of monitor 0x1 at (0,0)-(3840,2160), ignoring, maybe this is a problem with fullscreen.

Kkisak-valve maintainer 2023-05-08 github

The Witcher 3 (292030) Cannot Utilize CPU Clock Speeds

Issue transferred from https://github.com/ValveSoftware/Proton/issues/6752.
@Darkbloodking00 posted on 2023-05-08T09:39:20:

Compatibility Report

  • Name of the game with compatibility issues: The Witcher 3: Wild Hunt
  • Steam AppID of the game: 292030

System Information

I confirm:

  • [ Yes ] that I haven't found an existing compatibility report for this game.
  • [ Yes ] that I have checked whether there are updates for my system available.

ProtonLog: https://gist.github.com/Darkbloodking00/96858bfe0762fccfd06921df68abb31d

Symptoms

The Witcher 3 gradually lowers the CPU clock speeds until it hits 400-700 MHz which lowers my performance. I tried everything from changing the CPU governor to using different Proton versions. Even manually setting clock speeds and trying different BIOS options related to CPU power didn't fix my problem. I can confirm that it's not a heating problem since other games don't have such an unusual issue, my CPU can tolerate high temps like 93C, even if it's a heating issue clocks speeds shouldn't go below 400MHz.
witcher

Reproduction

1-Launch the game by any means either with DirectX11 or DirectX12 with our without launcher.
2-Set the graphics options to "LOW".
3-Continue your game or start a new one.
4-Either stay still or play the game normally a while.
5-Experience lower clocks speeds and lower performance.

Additional Information

Putting "WINE_CPU_TOPOLOGY=4:0,1,2,3 %command%" in launch options somehow fixes the problem but clock speed is still not in ideal range (it's locked to 2900 MHz) and peformance is low.


@zaps166 commented on 2023-05-08T11:55:58:

Do you use acpi-cpufreq with performance governor?
Did you change the power profile for integrated AMD GPU to LOW (idk why, but it can affect CPU speed)?
Are you sure you're not affected by SOC power limit? Is integrated GPU used somewhere else outside a game (it shares the same power limits with CPU)?
How about other games or other software?

Kkisak-valve maintainer 2023-05-12 github

Witcher 3 on Proton Experimental is not working.

Issue transferred from https://github.com/ValveSoftware/Proton/issues/6768.
@Bodyash posted on 2023-05-12T00:39:32:

Compatibility Report

System Information

  • GPU: AMD Radeon RX 6600M (navi23, LLVM 15.0.7, DRM 3.52, 6.3.1-arch2-1)
  • Driver/LLVM version: 4.6 Mesa 23.0.3
  • Kernel version: 6.3.1-arch2-1
  • Proton version: Experimental (8.0.2c)

I confirm:

  • [v] that I haven't found an existing compatibility report for this game.
  • [v] that I have checked whether there are updates for my system available.

Proton 8.0.2c logs:

https://gist.github.com/Bodyash/b869b1f20adc2e0c00b72934c8586607

Symptoms

So, I'm using "force use proton for unsupported titles" and set it to experimental. Witcher 3 is supported, but still it's defaults to Experimental. So game is not starting with this proton version, game launcher also doesn't work.
Using start params: gamemoderun mangohud %command% --launcher-skip
Current Proton Experimental version 8.0.2c
Also tried 7.0.6 - same symptoms.

And 8.0.2 is working fine, with no performance problems or any bugs.

Kkisak-valve maintainer 2023-05-12 github

Hello @Bodyash, it should be noted that Proton 8.0-2c is the current build of Proton 8.0, not Proton Experimental.

Please copy your system information from Steam (Steam -> Help -> System Information) and put it in a gist, then include a link to the gist in this issue report.

BBodyash 2023-05-12 github

@kisak-valve System Information https://gist.github.com/Bodyash/27331cfbc38ba47ecb83c0fb20aec55f

So yes, 8.0.2c - is current version (And It's working).
Experimental version is 1683293476 experimental-8.0-20230505

And experimental is not generating proton log with PROTON_LOG=1, just instantly crashing
Also I'm playing another game with Proton - it's Fallout 4. When switching to Experimental - symptoms are the same, game not starting, instant crash, no proton log file.

Probably it's not a Witcher 3 problem only.

https://github.com/ValveSoftware/Proton/assets/2454569/f0f3646d-b4ae-4dbc-bb52-7445e828a914

Kkisak-valve maintainer 2023-05-12 github

No Proton log hints that a snafu happened before Proton had a chance to run at all. If you completely close Steam, then run steam-runtime from a terminal, are there any hints in the terminal spew when you tell Steam to start the game with Proton Experimental?

BBodyash 2023-05-12 github

image
Seems like some file is missing

Kkisak-valve maintainer 2023-05-12 github

Yes, your Proton Experimental install is damaged or incomplete. Please search for Proton Experimental in your Steam library, go to the per-game properties of Proton, switch to Installed Files, and select Verify integrity of tool files.

Aalasky17 maintainer 2023-05-15 github

@zaps166

Regressions with Proton 8.0-1 (Game v4.02, DX11):

  • after alt-tab the game runs in windowed mode all the time (taskbar is visible). Changing the game settings (windowed/borderless/fullscreen) doesn't help. Also using WM's hotkey to go fullscreen doesn't help (looks like window size is 1px less than screen size). Only game restart helps. No problems with Proton 7.0-6.

Thank you for reporting this! It turns out that this regression is specific to xfce - we are investigating this :)

  • DualSense PS5 gamepad is not detected by the game when it's connected while game is already running. After reconnection with Proton 7.0-6 gamepad was working with vibrations, but no haptics.

We could not reproduce this one - could you please retest with Proton 8.0-2 and experimental-8.0-20230505 and check if you are still seeing this issue?

Zzaps166 2023-05-16 github

@alasky17

Thank you for reporting this! It turns out that this regression is specific to xfce - we are investigating this :)

Right, openbox always shows fullscreen properly :slightly_smiling_face:

We could not reproduce this one - could you please retest with Proton 8.0-2 and experimental-8.0-20230505 and check if you are still seeing this issue?

No difference, same issue (experimental and 8.0-2). I connect PS5 gamepad via USB, Steam Input is disabled. Maybe it might work sometimes by a chance, but in most cases gamepad doesn't work at all or works once for ~1s. I'm talking about reconnecting a gamepad or connecting it when game is already running.

ZZesko 2023-05-17 github

I remembered that Witcher 3 should work well a year ago with Steam and Proton 7 on Manjaro or Arch based distro for PC Desktop: AMD Ryzen 5800x and AMD Radeon RX 5700 and the last version of mesa.

/etc/environments shows:

AMD_VULKAN_ICD=RADV
VK_ICD_FILENAMES=/usr/share/vulkan/icd.d/radeon_icd.i686.json:/usr/share/vulkan/icd.d/radeon_icd.x86_64.json on steam

But today, I can not start Witcher 3 because:

Screenshot_20230517_132616

The some error messages show in the log:

  • Two modules/libraries msmpeg2vdec.dll and msvproc.dll are missing
[308:312:0517/132532.639:ERROR:dxva_video_decode_accelerator_win.cc(1399)] DXVAVDA fatal error: could not LoadLibrary: msmpeg2vdec.dll: Module not found. (0x7E)
[308:312:0517/132532.652:ERROR:dxva_video_decode_accelerator_win.cc(1407)] DXVAVDA fatal error: could not LoadLibrary: msvproc.dll: Module not found. (0x7E)
  • I tried to switch from Proton 7.0-6 to 8.0-2, the same issue, but there are other error messages:
    d3d12core.dll and msvproc.dll are missing:
Proton: Error while copying to "/home/zesko/Steam/steamapps/compatdata/292030/pfx/drive_c/windows/system32/d3d12core.dll": No such file or directory
Proton: Error while copying to "/home/zesko/Steam/steamapps/compatdata/292030/pfx/drive_c/windows/syswow64/d3d12core.dll": No such file or directory
...
[332:336:0517/133123.788:ERROR:dxva_video_decode_accelerator_win.cc(1407)] DXVAVDA fatal error: could not LoadLibrary: msvproc.dll: Modul nicht gefunden. (0x7E)

I tested Proton Experimental, still the same issue.

Mmbriar 2023-05-17 github

@Zesko post full proton log. Witcher 3 isn't the same game anymore than it was a year ago after the recent patches, but I suspect you have either a driver install issue or your wine prefix is busted somehow.

ZZesko 2023-05-17 github

Witcher 3 isn't the same game anymore than it was a year ago after the recent patches

Indeed

post full proton log.

here is the full proton log of Proton 7
steam-292030.log

Mmbriar 2023-05-17 github

From the log:

info: Required Vulkan extension VK_KHR_surface not supported

Yep, you don't have a working vulkan driver at all. I don't know why you're messing around with /etc/environment to override the driver, but i'd recommend to make sure vulkan-radeon lib32-vulkan-radeon are installed and then to uninstall amdvlk lib32-amdvlk vulkan-amdgpu-pro lib32-vulkan-amdgpu-pro (if any of those are installed).

ZZesko 2023-05-17 github

I do not use amdvlk lib32-amdvlk vulkan-amdgpu-pro lib32-vulkan-amdgpu-pro

Finally, I now know the problem is with the wrong config in /etc/environment:

VK_ICD_FILENAMES=/usr/share/vulkan/icd.d/radeon_icd.i686.json:/usr/share/vulkan/icd.d/radeon_icd.x86_64.json on steam

is wrong, when I followed Arch Wiki

The correct config is:

VK_ICD_FILENAMES=/usr/share/vulkan/icd.d/radeon_icd.i686.json:/usr/share/vulkan/icd.d/radeon_icd.x86_64.json

Witcher 3 can start now with Proton 7. Thanks!

Mmbriar 2023-05-17 github

If you do not have amdvlk or amdgpu-pro vulkan installed then you don't need the config in /etc/environment at all.

Kkisak-valve maintainer 2023-05-24 github

The Witcher 3: Missing sounds and voices and DX11 option doesn't work since 4.0.3 update

Issue transferred from https://github.com/ValveSoftware/Proton/issues/6796.
@coreybruce posted on 2023-05-24T02:31:00:

Compatibility Report

  • Name of the game with compatibility issues:
  • Steam AppID of the game: 292030

System Information

  • GPU: RX 6700XT
  • Driver/LLVM version: Mesa 23.0.3-1
  • Kernel version: 6.1.29-1-MANJARO
  • Proton version: 8.0.1

I confirm:

  • [x ] that I haven't found an existing compatibility report for this game.
  • [ x] that I have checked whether there are updates for my system available.

(Proton logs compress well if needed.)-->
DX12 log
https://gist.github.com/coreybruce/1f9fd6d210ed26abc35041d26da6f179

DX11 Log
https://gist.github.com/coreybruce/54ba2168db427ceadb5f05ab1e302d81

Symptoms

  • Sounds and voices do not work, people talking are mute and only loading video cut-scene has sound.
  • DX11 fails to start, only DX12 option works

Reproduction

  1. Play game using DX12 and proton 8.0.1, also tested this with experimental with the same results
  2. Try and use the DX11 option and the game will fail to run.
Aalasky17 maintainer 2023-06-26 github

Regressions with Proton 8.0-1 (Game v4.02, DX11):

  • after alt-tab the game runs in windowed mode all the time (taskbar is visible). Changing the game settings (windowed/borderless/fullscreen) doesn't help. Also using WM's hotkey to go fullscreen doesn't help (looks like window size is 1px less than screen size). Only game restart helps. No problems with Proton 7.0-6.

@zaps166 This should be fixed with the latest proton-experimental release. Please let me know if you see any lingering regressive behavior with xfce here :)

Zzaps166 2023-06-26 github

Replying to https://github.com/ValveSoftware/Proton/issues/607#issuecomment-1608065739

Yes it's working now, thanks :slightly_smiling_face:

But gamepad hot-plugging stil doesn't.

Ccoreybruce 2023-06-27 github

Please fix the game next, dx11 mode is completely broken and DX12 audio doesn't fully work making it almost unplayable

Replying to https://github.com/ValveSoftware/Proton/issues/607#issuecomment-1561023033

Zzaps166 2023-06-27 github

Please fix the game next, dx11 mode is completely broken and DX12 audio doesn't fully work making it almost unplayable

Replying to [#607 (comment)](https://github.com/ValveSoftware/Proton/issues/607#issuecomment-1561023033)

Strange, recently I didn't play except tests, but DX11 works to me, I have voices (Polish language).

Ccoreybruce 2023-06-28 github

Replying to https://github.com/ValveSoftware/Proton/issues/607#issuecomment-1609822252

That is very odd, I cleared the proton cache and even verified my files but most of the sounds are missing in DX12 and Dx11 mode on proton 8.0-1 and Experimental is broken

When I select Dx11 it crashes
image

Video showing the sound not working, only sound on the video cut scenes play
https://i.imgur.com/uQEWbLo.mp4

Nnonviable 2023-07-05 github

patch 4.03 broke the textures of the game. on 4.02 everything worked great
image

KK0bin 2023-07-05 github

@justsleeper VKD3D-Proton Version? Proton version? GPU? Driver version?

Nnonviable 2023-07-05 github

@K0bin
01:00.0 VGA compatible controller: NVIDIA Corporation TU116 [GeForce GTX 1660 SUPER] (rev a1)
Subsystem: ASUSTeK Computer Inc. TU116 [GeForce GTX 1660 SUPER]
Kernel driver in use: nvidia
OpenGL version string: 4.6.0 NVIDIA 530.41.03
steam compatibility set to "Proton Experimental".. 8.0-2 same problem

Zzaps166 2023-07-07 github

@alasky17 @kisak-valve Could you please look at gamepad (DS5) hot-plugging when Steam Input is disabled (USB and Bluetooth)? It also affects other games like Hogwarts Legacy. No issue with Proton 7.0. On Proton 8.0 and experimental can't connect gamepad after starting game and can't re-connect gamepad - games just don't detect it.

Aalasky17 maintainer 2023-07-10 github

@zaps166 We have tried to reproduce the hotplugging issue you mentioned several times on different machines with no luck. It is working just fine over here. Can you think of anything unique about your setup? Are you using any special controller drivers? Are you running Witcher 3 in dx11 or dx12 mode? Are you plugging the controller directly into your computer or using some intermediary hub etc? I am using the Sony PlayStation 5 DualSense Wireless Controller Model CFI-ZCT1W - is this what you have? When you are testing plugged in, are you just using a USB-A to USB-C cord or something else? Do you have Steam Input specifically disabled somewhere - if so, where (it is disabled for Witcher 3 by default too)?

Also - could you get a log of the hotplugging failure with PROTON_LOG="+hid,+hidp,+hid_report,+plugplay,+rawinput,+dinput,+xinput,+joystick,+setupapi,+input" %command%?

Zzaps166 2023-07-11 github

@alasky17

Steam Version: 1689034492
Steam Settings->Controller->Enable Steam Input for Playstation controllers - OFF
Game properties->Compatibility - Proton Experimental (Proton: 1688590094 experimental-8.0-20230705b)
Game properties->Controller->Override for The Witcher 3: Wild Hunt - Use default settings (same as Disable in this case)

Gamepad: Sony Corp. DualSense wireless controller (PS5) (CFI-ZCT1W, Made in China xD)

Are you using any special controller drivers?

Kernel: 6.1.38-1-MANJARO, no additional drivers.

Are you running Witcher 3 in dx11 or dx12 mode

dx11 (I can reproduce this on some other games, too like Hogwarts Legacy)

Are you plugging the controller directly into your computer or using some intermediary hub etc?

Directly USB-C->USB-A cable or Bluetooth (doesn't matter, same issue).

Also - could you get a log of the hotplugging failure with

Yes, but is it any way to upload logs for you without making them public?


When I go to "Game properties->Compatibility" and select Proton 7.0-6, hot plugging is working (I can connect/disconnect many times):

  • on Bluetooth I've got Xbox glyphs in Wither 3
  • on USB I've got Playstation glyphs in Wither 3
Aalasky17 maintainer 2023-07-12 github

@zaps166 There would be a catch-22 with uploading the logs in a private way because I would have to write out the instructions on here :) Instead, the solution we came up with is for you to encrypt the log and then we can encrypt it with a privatekey.

  1. Download https://hiler.eu/gpg.txt
  2. Encrypt the file like this: gpg --encrypt --armor --recipient-file gpg.txt steam-1234.log
  3. Upload log here (or if file extension is wrong, upload to https://www.codeweavers.com/xfer/alasky/ps5-hotplugging/ with lock code witcherhogwartsplug

Edit: Could you also upload a dmesg log (just the output of running dmesg)? We are hoping that might shed some light on what is causing the difference in behavior.

Zzaps166 2023-07-12 github

@alasky17 Thanks!

Here is encrypted tarball containing dmesg output and logs from proton-experimental and working proton 7.0-6 (.txt suffix added for github): logs-witcher3-gamepad.tar.asc.txt

Aalasky17 maintainer 2023-07-13 github

@zaps166 Thank you so much! I asked someone to take a look at the log, and something very weird is going on but the cause is still unclear. Could you try a fresh prefix to see if that changes the behavior? Instead of deleting the prefix, please just rename/move it so that we can examine it if this ends up fixing the issue :) You probably already know this but - prefix located by default at ~/.steam/steam/steamapps/compatdata/292030.

Ccoreybruce 2023-07-13 github

Can someone look into my issue next please? The game is basically unplayable without the sounds and dx11 doesn't work for me anymore

Zzaps166 2023-07-13 github

@coreybruce

Can someone look into my issue next please? The game is basically unplayable without the sounds and dx11 doesn't work for me anymore

Something is wrong with your setup. I've looked at your logs - you have RDNA2 like me, Manjaro 6.1.x kernel like me and Proton 8.0-2. With this setup I don't have any issues with DX11 and I have voices (4.0.3 game version, self-compiled Mesa 23.1.3).

Maybe verify game files, delete your wine prefix (~/.steam/steam/steamapps/compatdata/292030) and try again? Do you use ntfs3 or ntfs-3g for your NTFS partition?


info:    Memory Heap[2]: 
info:      Size: 256 MiB
info:      Flags: 0x1
info:      Memory Type[3]: Property Flags = 0x7
info:      Memory Type[4]: Property Flags = 0x7
info:      Memory Type[9]: Property Flags = 0xc7

Also try to enable "Above 4G Decoding" in UEFI.

Ccoreybruce 2023-07-13 github

Replying to https://github.com/ValveSoftware/Proton/issues/607#issuecomment-1634526746

None of my other games have this issue, only this one ever since they pushed out that latest update..

Hmm I see so why did this update broke the game for me I wonder? 🤔

I could delete the compatdata folder and see if that fixes it

EDIT: nope even letting it recreate the prefix doesn't fix the issue, sounds like voices are missing in dx12 and dx11 still is broken .

No you can't use NTFS because proton has some sort of issue with it and wouldn't use it for Linux, my drive is EXT4

I've got that enabled in my bios also

Zzaps166 2023-07-13 github

@alasky17

Could you try a fresh prefix to see if that changes the behavior?

No difference, doesn't work either.

I noticed that sometimes (very rare case) it might work or works for ~1 sec, but after another reconnection doesn't work at all.

Zzaps166 2023-07-13 github

No you can't use NTFS because proton has some sort of issue with it and wouldn't use it for Linux, my drive is EXT4

I use NTFS3 (kernel driver) without issues, but it's tricky (best to disable Steam shader cache (in the past it was destroying data there and required checkdisk) and create symlink to compatdata which is on EXT4 :smile:

It's also possible to use ExFAT, but compatdata must be mounted by mount -o bind.


Hmm I see so why did this update broke the game for me I wonder? thinking

If it's ntfs-3g (fuse) everything bad might happen - update is modifying data which can be corrupted by driver...

I've can see 2023-05-24 12:22:56.781 [Information][ (0)] [TID 436][red-updater]: Current working directory: 'C:/users/steamuser/AppData/Local/Programs/CD Projekt Red/REDlauncher', filesystem: 'NTFS'. in your logs, so I thought you have NTFS.


nope even letting it recreate the prefix doesn't fix the issue, sounds like voices are missing in dx12 and dx11 still is broken .

Verify game files + manually check there aren't redundant files there.

I've got that enabled in my bios also

You should have 2 heaps, 3rd with 256 MiB shouldn't be visible. For me "Above 4G Decoding" is enough for Linux, but for real Windows I have to enable Re-bar, too (AMD Ryzen).

BBlisto91 2023-07-13 github

@coreybruce I assume you don't have any old mods or mod files in your game folder?

I also see a reddit thread post about someone who had a similar audio issue. There the solution was to delete some game patch folders. Tho that was for the next gen update and not 4.03 specifically https://www.reddit.com/r/Witcher3/comments/12ysjsk/fix_for_the_no_voice_audiomusic_audio_bug_after/

I haven't personally had these issues myself.

Aalasky17 maintainer 2023-07-13 github

@zaps166 After further investigation, we suspect something related to your setup is causing the difference. Could you try a fresh game install (with clean prefix) on non-ntfs?

Zzaps166 2023-07-13 github

@alasky17 I'll check later on EXT4, but I can reproduce it on many games, just checked:

  • Witcher 3
  • CP2077
  • Stray
  • Hogwarts Legacy

and probably much more... Tested games switched to Proton 7.x are ok with hot plugging.

Just set this per every game and compare 7.0-6 vs 8.0-2 or Experimental:
Screenshot_20230713_230818

Zzaps166 2023-07-13 github

@alasky17

Tried EXT4 (with clean prefix) and linux 6.5rc1, both also doesn't work.

But I noticed a difference comparing to NTFS3 with symlinked compatdata (maybe it was a coincidence):

  • on EXT4 I can hotplug 0x-3x (usually 0x-1x)
  • on NTFS3 I can hotplug 0x-1x (usually 0x)

When connecting a gamepad, after a while I can see short game freeze (looks like ~250ms) even if gamepad is not working in game.


How many times did you try to connect/disconnect gamepad?

Aalasky17 maintainer 2023-07-13 github

@zaps166 Just to confirm - when you tested EXT4, that was a "standard" test without any custom symlinks in the process, right?

I've been testing only with the USB cord (not bluetooth) because that is simpler. I was testing only a couple of times plugging and unplugging, but this time I went all in - I did 10+ times hotplugging, I tried starting the game with the controller disconnected, I tried hotplugging while in-game instead of the main menu, no luck. I double checked that I have steam input disabled. It sounds like none of this should be necessary (aside from trying at least 5x because sometimes the first hotplug works) since you can just launch the game and then see the problem, even on the menu?

Do you have any other peripherals (like a wheel, VR headset etc) plugged in? If so, could you try unplugging anything extra just in case that is related? (Surprisingly, this has been the problem more than once in the past). Do you have a special monitor array, etc?

Ccoreybruce 2023-07-14 github

I just re downloaded the game and and it's still doing it 😞

no other games are this issue other and nah I don't have any mods for this games and haven't modded it

Ccoreybruce 2023-07-14 github

tested the game on my gaming laptop with a Nvidia 166o TI using driver 535.54.03 on Manjaro and the menu music and sounds including voices work. Is this a issue mesa issue?

I have mesa 23.0.4-2 on my desktop

Zzaps166 2023-07-14 github

@coreybruce

Is this a issue mesa issue?

I have mesa 23.1.3, tried also 23.0.4-2 from Manjaro stable repo - no issues for me, sound is working, DX11 is working.

sounds like voices are missing

Witcher3 has 2 audio outputs, one is for cutscenes. Are you sure one of them it isn't muted? Check it in pavucontrol when game is open.


Maybe disable Steam shader cache in Steam settings, remove ~/.cache/mesa_shader_cache, remove witcher3.dxvk-cache and vkd3d-proton.cache files from Witcher 3 game bin directory, check if you don't have any weird global and game envars set like SDL_AUDIODRIVER or RADV_PERFTEST.

Ccoreybruce 2023-07-14 github

Replying to https://github.com/ValveSoftware/Proton/issues/607#issuecomment-1635373317

Hmm I see, weird

No no like I said I have sound but some sounds are missing like the menu music and voices for example but cut scenes and other sounds work.

Why would shader cache have anything to do with audio issues or isthis to do with DX11 not working?

I have no launch options on this game either.

Zzaps166 2023-07-14 github

@alasky17

Re-downloaded Proton-Experimental, unmounted NTFS data (no NTFS library, no symlinks), single game in library, removed entire compatdata, I used USB to test, I'm in game menu.

Do you have any other peripherals

Mouse, keyboard, WiFi (unused, using ethernet), Bluetooth (disabled for now by rfkill), ASUSTek Computer, Inc. AURA LED Controller (not possible to disable this ****, it's built-in in motherboard...).

Do you have a special monitor array, etc?

Single monitor 3840x2160, Xorg, full screen with compositor bypassed, FreeSync is working.


Steam logs from terminal when connecting gamepad:

Local Device Found
  type: 054c 0ce6
  path: sdl://5
  serial_number:  - 0
  Manufacturer: 
  Product:      DualSense Wireless Controller
  Release:      100
  Interface:    -1

!! Steam controller device opened for index 0.
!! Controller 0 attributes:
  Type: 45
  ProductID: 3302
  Serial: [removed by me]
  Capabilities: 02975bff
  Firmware Version: 854
  Firmware Build Time: 2147483647 (Tue, 19 Jan 2038 03:14:07 GMT)
  Bootloader Build Time: 2147483647 (Tue, 19 Jan 2038 03:14:07 GMT)
Loaded Config for Local Selection Path for App ID 413090, Controller 0: controller_base/bigpicture_gamepad.vdf
Loaded Config for Local Selection Path for App ID 413090, Controller 0: controller_base/bigpicture_gamepad.vdf
Loaded Config for Local Selection Path for App ID 769, Controller 0: controller_base/basicui_gamepad.vdf
Loaded Config for Last Resort Path for App ID 769, Controller 0: /home/user/.local/share/Steam//controller_base/basicui_gamepad.vdf
Loaded Config for Local Override Path for App ID 413080, Controller 0: /home/user/.local/share/Steam//controller_base/empty.vdf
Loaded Config for Local Override Path for App ID 413080, Controller 0: /home/user/.local/share/Steam//controller_base/empty.vdf
Loaded Config for Local Override Path for App ID 413080, Controller 0: /home/user/.local/share/Steam//controller_base/empty.vdf

I'll check second PC later.

Zzaps166 2023-07-14 github

@coreybruce

Why would shader cache have anything to do with audio issues or isthis to do with DX11 not working?

Shader cache for DX11. It's strange, I've got same system with RDNA2 GPU and it's working to me.

Ccoreybruce 2023-07-14 github

@coreybruce

Why would shader cache have anything to do with audio issues or isthis to do with DX11 not working?

Shader cache for DX11. It's strange, I've got same system with RDNA2 GPU and it's working to me.

Yeah it's very odd, after I deleted that cache file the game wanted to do this 4.3mb update
image

Also game refuses to run on DX11 using proton 8.0-2

Ccoreybruce 2023-07-14 github

Nice the launcher is now black... They shouldn't have releases this update, f#ck sakes..

I know you can skip this launcher but I wanted to test if DX11 etc is working properly
image

Nnetfab 2023-07-14 github

I know you can skip this launcher but I wanted to test if DX11 etc is working properly

I'm getting also the black launcher problem since some time.
You can skip the launcher with --launcher-skip option on command line.
And if you want to run with DX11, in the game installation directory you will find the following file : launcher-configuration.json
You can edit this file and set the fallback line to "DirectX 11".

Aalasky17 maintainer 2023-07-14 github

WRT the black launcher - it should be fixed in 8.0-3 (currently still a beta RC on 8.0-2) but if you want to test 8.0-2, you can use LIBGL_ALWAYS_SOFTWARE=1 %command% in the launch options :)

@zaps166 Thank you for the additional info. I'll try to keep digging, but I might be stumped until someone comes back from vacation which will be another full week. I wanted to give you a heads-up on the timeline for fresh ideas :)

Zzaps166 2023-07-14 github

@alasky17 Same on second PC. Once it worked, but PS5 gamepad was detected as Xbox. On next game restart, hot plugging stopped working after 2 re-connections. Another game restart - hot plugging didn't worked at all.

HHooloovoo 2023-07-16 github

Hello,

Am I understanding correctly that you want one GitHub issue for all issues with the same game? If so, I am seeing a regression between Proton versions on Witcher 3 with all else being the same.

I have a NVIDIA Corporation GK104 [GeForce GTX 770] and Intel® Core™ i7-6700 CPU @ 3.40GHz × 8, so it is a relatively old machine and cannot do DirectX 12, but it can do DirectX 11.

Using Proton 7.0-6 I can play, adding the "--launcher-skip" command to launch options and changing the launcher-configuration.json entry for "fallback": "DirectX 11".

If I change the Compatibility settings to use Proton 8.0-2 or Proton Experimental, it gives an error:
image
"GPU does not meet minimal requrements. Support for DirectX 11 is required."

If I change it back to Proton 7.0-6 then it works again.

Running on Ubuntu 22.04.

Please let me know if there are any logs etc that it would be helpful for me to collect.

Operating System Version:
Ubuntu 22.04.2 LTS (64 bit)
Kernel Name: Linux
Kernel Version: 5.19.0-46-generic
X Server Vendor: The X.Org Foundation
X Server Release: 12101004
X Window Manager: GNOME Shell
Steam Runtime Version: steam-runtime_0.20230509.49499
Video Card:
Driver: NVIDIA Corporation NVIDIA GeForce GTX 770/PCIe/SSE2
Driver Version: 4.6.0 NVIDIA 470.199.02
OpenGL Version: 4.6

BBlisto91 2023-07-16 github

@Hooloovoo Hi there.
Since Proton 8 the Vulkan driver requirements have been upped which means for Nvidia a minimum driver version of 510 is needed. This sadly means that kepler series cards will have to stick to Proton 7

Aalxwk 2023-07-21 github

WRT the black launcher - it should be fixed in 8.0-3 (currently still a beta RC on 8.0-2) but if you want to test 8.0-2, you can use LIBGL_ALWAYS_SOFTWARE=1 %command% in the launch options :)

Neither of those (setting the variable or using 8.0-3 beta) fixes the black launcher issue for me. I tried to delete the prefix and everything. Before the update the deletion of the prefix helped to launch it at least once (after the first launch the black window returned), but now even the first launch went black. Maybe that's something related to issues with using vaapi on Nvidia (I'm using 535.86.05 proprietary drivers).

Kkisak-valve maintainer 2023-07-21 github

Hello @alxwk, please add PROTON_LOG=1 %command% to the game's launch options and attach the generated $HOME/steam-$APPID.log to this issue report as a file. (Proton logs compress well if needed.) Also, please copy the contents of Steam Runtime Diagnostics from Steam (Steam -> Help -> Steam Runtime Diagnostics) and put it in a gist, then include a link to the gist in this issue report.

Rrbernon maintainer 2023-07-27 github

Hi @zaps166, we tried to reproduce the hotplug issue you are experiencing, but to no avail so far. I have uploaded a custom Proton - Experimental build to a "test-bug-22500" branch (which you can select in Steam, in the Proton - Experimental tool properties, under the Betas tab), with some possible culprit changes reverted. Would you mind trying it and see if it solves your problem?

Edit: Also, if you have some other controller models, I would be interested to know if they suffer from the same issue or if it seems to really be PS5 controller specific.

Zzaps166 2023-07-30 github

@rbernon It seems to be working now, thanks! :grin:


I don't have other controllers, I can check DS5 only now.


The only problem is that after hot plugging game sometimes detects gamepad correctly as PS5, but many times detects as XBox gamepad. But it's a different issues - I can see similar behavior on Proton 7.0-6 (but in Hogwarts Legacy in 7.0-6 after hot plugging it detects as PlayStation gamepad with vibrations with different button for map hotkey :rofl:).

Generally hot plugging is working now! But there are still (other) issues + Hogwart's Legacy is a special case, because game doesn't support DS5 gamepad correctly even on Windows AFAIK.

Rrbernon maintainer 2023-07-31 github

Thanks for the confirmation. The detection as an XBox gamepad is weird but looks like a different issue.

Ccoreybruce 2023-08-01 github

Has there been any progress on the audio and launcher problem?

Aalasky17 maintainer 2023-08-01 github

@coreybruce The launcher should no longer be black with 8.0-3, but if you are testing older versions you can use LIBGL_ALWAYS_SOFTWARE=1 %command%. As for the audio issue, as far as I can tell from this thread, there is something going wrong specific to your setup :/ If the audio used to work, I think the best way to move forward would be to figure out which thing updating broke it.

Ccoreybruce 2023-08-02 github

@coreybruce The launcher should no longer be black with 8.0-3, but if you are testing older versions you can use LIBGL_ALWAYS_SOFTWARE=1 %command%. As for the audio issue, as far as I can tell from this thread, there is something going wrong specific to your setup :/ If the audio used to work, I think the best way to move forward would be to figure out which thing updating broke it.

Yeah looks like the launcher is now working correctly and I am using the latest 8.0-3 but I am still running into the issue of dx11 not working and sound not working in dx12 sadly.

The odd thing is I'm not having this issue with any other game except this one ever since they brought out this update which started this problem.. I've been playing Ratchet and Clank, Spiderman, Forza Horizon 5 and many other games with no issues (except Red Dead Redemption 2 which could be related to a mesa issue which causes Wayland to crash when you got to a certain area) but other than that been playing games with sound and running fine.

Right now I'm going to test out the classic 1.32 version in the betas and see if it works. Really wish they had a option to roll back to the last version before this problematic update.

Ccoreybruce 2023-08-02 github

So on classic 1.32 patch with proton 8.0.3 DX11 works and I even have all the sounds in the game working including the menu music but if I try DX 12 it will say in the launcher that it failed to launcher and to verify the files which I did and even tried experimental but it still does it but of course I can't load my save because it' has been updated for the latest version..

image

Here is the log: https://gist.github.com/coreybruce/f46d4cb010fc5d55af03ad8fd6e142a4

I am using a RX 6700XT with mesa 23.0.4-3 currently

This game is very problematic now

Aalasky17 maintainer 2023-08-09 github

@zaps166 The reverts that you tested with the custom build are now live in Proton - Experimental. Please let us know if you are still seeing any hotplugging issues with experimental or if the regression is resolved 100%.

Zzaps166 2023-08-15 github

@alasky17 Works on Proton-Experimental, thanks! But when hot-plugging, it's mostly detected as XBox, sometimes as PS5.

Aalasky17 maintainer 2023-08-15 github

@zaps166 Good to hear! Is this any different from 7.0-x behavior? Many games don't have the best support for PS5 ...

Zzaps166 2023-08-21 github

@alasky17 I think Witcher 3 is working the same now as 7.0-x :smile: In 7.0-x Hogwarts Legacy was detected as PlayStation3 (I guess, because different button was used to open a map in game) instead of XBox on hot-plugging, but this game supports DS5 poorly anyway.
But detecting as XBox quite often (not always) is strange on these both games.

Zzaps166 2023-09-24 github

Same as https://github.com/ValveSoftware/Proton/issues/6510#issuecomment-1723920995, game freezes when monitor/TV is turned off/on or unplugged/plugged-in again. Happens on windowed, borderless, fullscreen. Here happens the same but without EXCEPTION_INT_DIVIDE_BY_ZERO. Sometimes can't press STOP on Steam because mouse is still grabbed (anyway is it a hotkey for this?).

steam-292030.log.txt

Zzaps166 2023-10-01 github

This should be fixed with the latest proton-experimental release. Please let me know if you see any lingering regressive behavior with xfce here :)

@alasky17 This problem is back on bleeding-edge (and experimental, too) :disappointed: Is it really Xfwm4 bug, because other full screen apps (non-proton) are working properly. I can try to fix it on Xfwm4 side if wine can't workaround it, but I need more information what's exactly wine is doing with the window.

Aalasky17 maintainer 2023-10-02 github

@zaps166 This is an extremely complicated area :/. There is some behavior in Wine that could be considered "buggy" and some behavior in XFCE that could be considered "buggy". The dev who was working on this actually just got a fix committed to xfwm4 for this bug. https://gitlab.xfce.org/xfce/xfwm4/-/commit/215999ab15a310e98f3fb46118018d23a2ccd62d. It would be great if you could try xfwm4 with this commit and see if you still are seeing issues.

Zzaps166 2023-10-02 github

@alasky17 Thanks! I've applied the patch, full screen is working correctly , but borderless doesn't allow to alt-tab (it's like always on top) :disappointed: Proton 7.0-6 and (IIRC) previous proton-experimental allowed me to alt-tab the borderless Witcher 3. EDIT: Re-checked, it's working - strange.

Aalasky17 maintainer 2023-10-03 github

@zaps166 Awesome! Please let me know if you find any other regressions ... we can't always fix window behavior, but we try to not go backwards :)

Zzaps166 2023-10-03 github

@alasky17 Window behavior is correct on Xfwm4 with the patch! :slightly_smiling_face:

Please let me know if you find any other regressions

Two regressions I found so far (tested on various games):

  • since Proton 8.0 disconnecting a monitor (or if you turn monitor/TV off and on) game freezes and needs to be killed.
  • there are still issues on proton-experimental and experimental-bleeding-edge-8.0-57984-20231003-pdf5240-w7a90e2-d494f7f-v4bf519 with a mouse grab on game exit (proton-8.0-4-rc7 is OK in this case).
Aalasky17 maintainer 2023-10-04 github

@zaps166

  • I was able to reproduce the monitor disconnect issue with Witcher 3 and we will look into it :)
  • wrt the mouse being captured on game exit, this should be fixed with current bleeding-edge (fix is not in regular experimental-0929b). In my testing, the cursor now is released the same as 8.0-3 (although of course some games still don't like to let it go). If you are still seeing the issue with experimental-BE, please comment on a game issue where you see the regressive behavior and tag me in the comment, so that we can keep this issue strictly for Witcher 3. Thank you for the follow-up!
Zzaps166 2023-10-04 github

@alasky17

Thanks :)

I have experimental-bleeding-edge-8.0-58166-20231004-pdf5240-w513d37-d494f7f-ve227e9 - is it the newest?

Edit: I can see randomness with cursor ungrab on one game (sometimes ok, sometimes not ok), other game is always not ok. Please let me know if this BE contains the patch - if yes, I'll ping you in these games :slightly_smiling_face:

Aalasky17 maintainer 2023-10-05 github

@zaps166 yes, that experimental should be new enough. Please ping me in the other games :)

Aalasky17 maintainer 2023-10-12 github

@zaps166

  • since Proton 8.0 disconnecting a monitor (or if you turn monitor/TV off and on) game freezes and needs to be killed.

Could you please test experimental bleeding-edge and see if this is fixed for you now? It would be great to test multiple games where you saw this failure, just in case the fix is only satisfying some games :)

Zzaps166 2023-10-13 github

@alasky17 Thanks! I'ts working, also I noticed mouse cursor grab fix on exit :slightly_smiling_face:

But I think Xfwm4 has another bug, because some games freezes or running very slow due to very large windows size and causes huge system RAM consumption. On Openbox it works OK. Could you look if it's really an Xfwm4 issue that it creates large windows like 22093 or 32767?

Edit: Xfwm4 can use uninitizlized values as full screen window size in some cases, I created MR.

NNewbytee 2023-11-08 github

The game crashes for me with the "Damn, that's ugly" popup when I try to launch it in DirectX 11 mode. DirectX 12 mode works, but performance leaves much to be desired. Log: steam-292030.log

I've tried recreating the Wineprefix, verifying the game's files, and skipping the launcher (skipping the launcher not in effect in the log above). This is with the latest Proton Experimental. Before the big update I was able to play the game just fine.

GPU: GTX 970
Driver version: 535.129.03
Distribution: Flathub/Freedesktop runtime 23.08

Zzaps166 2023-12-02 github

@alasky17

It's not a regression, but could you look at mouse grab once again?

I have single monitor (Xorg) and play games with borderless fullscreen option - the window fills the entire screen, so it's like exclusive full screen on Windows. The problem is that mouse is grabbed into a game in this case. If I'll get a popoup (for example system notification or new message notification) or any window with BypassWindowManagerHint flag, I can't click it even if game shows system cursor - I have to press Alt-Tab which is a bit annoying.

It affects many games (can behave differently), I'll give steps to reproduce in Witcher 3 game:

  • use single monitor
  • go to Options->Video->Display, set "Borderless Window" and enable Hardware Cursor
  • get a notification message while you're in game menu (when you have free cursor) and try to click it - you'll click in the game even if mouse cursor is over the popup window

Is it a quick way to disable mouse grabbing on full screen?

Eebouther 2023-12-07 github

Game often freeze when navigating in map menu for me. Audio continues to play.
I'm using proton experimental.
Here's proton logs:

47374.047:03ec:0470:fixme:secur32:get_mac_algid unknown algorithm 200, cipher 23
47388.138:03ec:0494:warn:vkd3d-proton:d3d12_fence_set_native_sync_handle_on_completion_explicit: Event completion for native sync handle is already in the list.
47403.953:0100:0110:fixme:oleacc:find_class_data unhandled window class: L"[#32769](/issue/ValveSoftware/Proton/32769)"
47403.953:0100:0110:fixme:uiautomation:uia_get_providers_for_hwnd Override provider callback currently unimplemented.
47403.953:0100:0110:fixme:uiautomation:msaa_provider_GetPropertyValue Unimplemented propertyId 30024
47403.954:0100:0110:fixme:uiautomation:uia_element_GetCurrentPropertyValueEx Default property values currently unimplemented
47403.954:0100:0110:fixme:uiautomation:uia_element_GetCurrentPropertyValueEx Default property values currently unimplemented
47403.954:0100:0110:fixme:uiautomation:msaa_provider_GetPatternProvider Unimplemented patternId 10002
47403.954:0100:0110:fixme:uiautomation:nc_provider_GetPatternProvider 00000000008DEEC0, 10002, 0000000001B2EC90: stub
47403.954:0100:0110:fixme:uiautomation:base_hwnd_provider_GetPatternProvider 00000000008DEE30, 10002, 0000000001B2EC90: stub
MMohandevir 2024-01-24 github

Replying to https://github.com/ValveSoftware/Proton/issues/607#issuecomment-1845773587

Since Proton 8.0.5 release, I'm unable to play Witcher 3 with DX12. The game constantly freezes after a couple of minutes, if not instantly, when trying to access the glossary sub menus. DX11 runs fine. Tested on my Steam Deck and my PC (Ryzen 5 3600 & RX 6600) with Holoiso. Same behavior. I tried deleting the proton files and validating the installation files, but it didn't solve the issue.

Edit: For the record, I was left unable to read the Red Miasmal bestiary entry for the "In the Enternal Fire's Shadow" questline, which coincides with Proton 8.0.5 release, in my run. Unfortunately it happened again when I tried to craft at Crow's Perch or any other armorer, after that.

Kkisak-valve maintainer 2024-01-24 github

The Witcher 3: Wild Hunt (292030): Glossary Bug

Issue transferred from https://github.com/ValveSoftware/Proton/issues/7440.
@matheusfaustino posted on 2024-01-24T16:58:41:

Compatibility Report

  • The Witcher 3: Wild Hunt (v4.0.4)
  • 292030

System Information

  • Steam Deck OLED
  • Stock SteamOS 3.5.7 stable
  • Experimental+Proton 8.0-5

I confirm:

  • [x] that I haven't found an existing compatibility report for this game.
  • [x] that I have checked whether there are updates for my system available.

steam-292030.zip

Symptoms

Every time I enter the Glossary submenu (if enter the Glossary through the other options, it works, the problem is when it shows the submenu and I try to interact with it) using DX12, which allows HDR on the OLED, the game freezes and I have to force-quit the game. This issue also occurs in Gwen's tutorial. I initially tried DX12 for HDR since I couldn't enable HDR for DX11. It only occurs via DX12.

Reproduction

  1. Launch the game, as usual, and select DX12.
  2. It happens to me, independently, of the graphic preset, but I am using the Steam Deck preset.
  3. Press Start (that 3 lines button on steam deck), it will show that stylish menu with Map/Quests/Inventory/etc
  4. Go to Glossary and press down, it will move your cursor to the submenu which contains (Bestiary/Tutorial/Characters/Books/Crafting)
  • After the submenu's animation, it freezes, and I cannot move anything (Even MangoHud is frozen).
  1. It can also be stuck like that, if I enter at Gwen's Tutorial card game (the first time you talk with the scholar and he teaches you about the card game)
BbastiHST90 2024-01-28 github

I'm on the steam deck OLED (stable steam client and os) and since the proton 8.0-5 update I have the same issue above.
It freezes in the menu. And not only glossary, even in the inventory or character screen.
Sound is still working and game can be closed with the steam menu.
I had 0 issues in 40h of playtime on proton 8.0-4

For now I use protonGE 8-2 (I still had that installed) which doesn't have this problem, but gave me 2 random game crashes (or more like OS crashes) in ~10h. The newest ProtonGE has the same issue as Proton 8.0-5.

Kkisak-valve maintainer 2024-01-28 github

Hello @1basti1, please add PROTON_LOG=1 %command% to the game's launch options, reproduce the regression, and attach the generated $HOME/steam-$APPID.log to this issue report as a file. (Proton logs compress well if needed.) EDIT: Never mind, https://github.com/ValveSoftware/Proton/issues/607#issuecomment-1908703903 has a relevant log.

BBlisto91 2024-02-05 github

@Mohandevir @matheusfaustino @1basti1 Can you think of any more specifics that might help us reproduce this freeze issue since 8.0-5? We've tried to reproduce it on a bunch of different setups including a OLED Deck but without luck so far.

Could you attach the save games this issue occurs with if they are tied to one? Though i assume not since the Gwent tutorial is pretty early in the game

BbastiHST90 2024-02-05 github

I could make a video in 3-4 hours.
But essentially it's exactly like described in this post

https://github.com/ValveSoftware/Proton/issues/607#issuecomment-1908703903

BbastiHST90 2024-02-05 github

@Blisto91
I hope streamable is ok. Video to big for GitHub

https://streamable.com/r5599n

I played with ProtonGE (older one because newest has the same freezing problem) for the last days without problems. Using proton 8.0-5 it immediately freezes. As you can see in the video.

I mean I use some decky plugins. But that wasn't a problem before and with protonGE it works flawlessly. I have no idea what the problem is.

I now have 71 hours on this game. When proton updated to 8.0-5 I was on ~40h.

Edit: Seeing the edit below me. I use a 512GB OLED deck which I upgraded to 1TB WD SN530. But that shouldn't matter.

MMohandevir 2024-02-05 github

I tried all my saves and it's the same (even my 5 years old ones). I even
completely reinstalled the game. No luck.

Press the menu button, go to glossary, press the down arrow... Freeze. I have to kill it through the Steam menu.

I will try to start a new game and see if it freezes.

Edit: I tried a new save game. It's the same behavior. There are no issues with DX12 on any saves, while in desktop mode, though.

Edit 2: Seeing the above video, I tried switching to english (from french, game and system), back and forth. Still the same, no matter the language.

Edit 3: Just loaded Proton-8.4-GE. Problem gone.

For the record, I have an original LCD 512Gb Steam Deck. I upgraded the nvme with a Sabrent Rocket 1tb.

Le lun. 5 févr. 2024, 7 h 01 a.m., Blisto91 @.***> a
écrit :

@Mohandevir https://github.com/Mohandevir @matheusfaustino
https://github.com/matheusfaustino @1basti1 https://github.com/1basti1
Can you think of any more specifics that might help us reproduce this
freeze issue since 8.0-5? We've tried to reproduce it on a bunch of
different setups including a OLED Deck but without luck so far.

Could you attach the save games this issue occurs with if they are tied to
one? Though i assume not since the Gwent tutorial is pretty early in the
game

—
Reply to this email directly, view it on GitHub
https://github.com/ValveSoftware/Proton/issues/607#issuecomment-1926834201,
or unsubscribe
https://github.com/notifications/unsubscribe-auth/ABQFRY4BVRUI67U5YFH7SNTYSDC2DAVCNFSM4FRSLUZ2U5DIOJSWCZC7NNSXTN2JONZXKZKDN5WW2ZLOOQ5TCOJSGY4DGNBSGAYQ
.
You are receiving this because you were mentioned.Message ID:
@.***>

BBlisto91 2024-02-06 github

@Mohandevir Are you also using Decky or is the nvme the only thing changed from a out of the box Deck? Besides Proton version that is.

MMohandevir 2024-02-06 github

Tought about that too... I was using Decky but I completely removed it some
time after SteamOS 3.5 release. I was usind the "Vibrant color" (?) plugin.
Do you think it could have left some "damaging" traces?

Le mar. 6 févr. 2024, 2 h 41 a.m., Blisto91 @.***> a
écrit :

@Mohandevir https://github.com/Mohandevir Are you also using Decky or
is the nvme the only thing changed from a out of the box Deck?

—
Reply to this email directly, view it on GitHub
https://github.com/ValveSoftware/Proton/issues/607#issuecomment-1928942456,
or unsubscribe
https://github.com/notifications/unsubscribe-auth/ABQFRY7Q46RAIBKK2CN5LGLYSHND7AVCNFSM4FRSLUZ2U5DIOJSWCZC7NNSXTN2JONZXKZKDN5WW2ZLOOQ5TCOJSHA4TIMRUGU3A
.
You are receiving this because you were mentioned.Message ID:
@.***>

BbastiHST90 2024-02-06 github

To add to that, I use cryo utilities recommended settings, but with a 20gb swap file (I know it's large, didn't remember why I set it to that, but oh well)

I really hope I can still use decky + cryo with later proton versions... I don't want to miss these things.

But Vram in bios is still at 1GB.

Because I don't play other games right now and I'm afraid other games might have the same issues...?

MMohandevir 2024-02-06 github

Yeah... Forgot about that. I use cryoutilities with an 8gb swap file, too.

Le mar. 6 févr. 2024, 12 h 41 p.m., 1basti1 @.***> a
écrit :

To add to that, I use cryo utilities recommended settings, but with a 20gb
swap file (I know it's large, didn't remember why I set it to that, but oh
well)

I really hope I can still use decky + cryo with later proton versions... I
don't want to miss these things.

But Vram in bios is still at 1GB.

—
Reply to this email directly, view it on GitHub
https://github.com/ValveSoftware/Proton/issues/607#issuecomment-1930450497,
or unsubscribe
https://github.com/notifications/unsubscribe-auth/ABQFRY4BMFP5F2XQ6FAHTN3YSJTK3AVCNFSM4FRSLUZ2U5DIOJSWCZC7NNSXTN2JONZXKZKDN5WW2ZLOOQ5TCOJTGA2DKMBUHE3Q
.
You are receiving this because you were mentioned.Message ID:
@.***>

Mmatheusfaustino 2024-02-06 github

Me too, using cryoutilities with default settings (16gb). But, my VRAM is at 4GB.

BbastiHST90 2024-02-06 github

ok, I think one of us really should test with original settings.
I can maybe do that tommorow.

Mmatheusfaustino 2024-02-06 github

@1basti1 @Mohandevir just tested it now, I return to the stock setup, with VRAM at 1GB, and it still failed at the same way. If you want to check as well, feel free. I am running everything stock now.

BbastiHST90 2024-02-06 github

@matheusfaustino Do you use decky loader?

Mmatheusfaustino 2024-02-06 github

@1basti1 decky loader, yes. With a lot of add-ons, but none that affects the game directly. This one will be tricky to uninstall. I could try, but I need to find a way to save my settings before.

BbastiHST90 2024-02-06 github

I think it's enough to disable cef debugging or something in developer settings. then it doesn't load after restart.
I'll test tomorrow if I don't forget

Mmatheusfaustino 2024-02-06 github

Just disabled the cef debugging and restarted (no decky loader in the menu) and it's still freezing for me, right after when I go to the submenu from Glossary :(

BbastiHST90 2024-02-06 github

Yeah, then I don't know... :(

MMohandevir 2024-02-06 github

If I didn't have so many games installed and all my emulator custom setups,
I would try a factory reset...

Le mar. 6 févr. 2024, 2 h 31 p.m., 1basti1 @.***> a
écrit :

Yeah, then I don't know... :(

—
Reply to this email directly, view it on GitHub
https://github.com/ValveSoftware/Proton/issues/607#issuecomment-1930617540,
or unsubscribe
https://github.com/notifications/unsubscribe-auth/ABQFRY5RNUBTDINEXC6EQJ3YSKAHRAVCNFSM4FRSLUZ2U5DIOJSWCZC7NNSXTN2JONZXKZKDN5WW2ZLOOQ5TCOJTGA3DCNZVGQYA
.
You are receiving this because you were mentioned.Message ID:
@.***>

Mmoonman5 2024-02-06 github

Hi, I have exactly the same symptoms as 1basti1, Mohandevir and matheusfaustino.
I have Steam Deck OLED 1TB, I have neither Decky nor Cryoutilities installed, no other plugins or system modifications. Everything is up-to-date and on stable channel. I could say it's "out of the box".
The only thing I did right before updating to proton 8.0-5 was installing Java to play Minecraft.

I tried to delete the proton 8.0 files along with its appmanifest file and re-download the proton, but it did nothing.

BbastiHST90 2024-02-10 github

Just fyi I uninstalled the game because I finished it. needed space. So I cant test anything anymore. But still hope it gets fixed.

Nnshank63 2024-02-13 github

I have the same issue. Is there a confirmed fix for this yet? I don't really want to play on DX11

I will try some of the proton versions mentioned above after work and report back....but I tested multiple versions last night and the game froze in menu regardless of proton version

OLED, DX12, stable channel. I think I tried the lastest stable Proton, Proton Experimental, and Proton GE 31? I think that was it.

BBlisto91 2024-02-13 github

@nshank63 hi there and thank you for the report.
You mean that the game can freeze up during certain game actions like opening the bestiary?

Have you made any modifications to your Deck, software (Decky etc) or hardware wise?
We are still trying to track down a reproduction on this issue.

Nnshank63 2024-02-13 github

@Blisto91 Correct. Anytime I try to access that menu or any submenus like alchemy...it freezes up.

It's an OLED I got at launch and installed the 1TB SSD from my LCD deck in the day I got it. I have an 8GB swap file so RDR2 wouldn't crash after X hours. No other modifications. No issues with any other games or even the DX11 version of this game. DX12 is so much better though I would love to use it if we can figure out a fix.

The video the user posted above sums up what happens perfectly. I have the exact same issue.

I have tried a bunch of different proton versions. reinstalling the game, validating files etc.

I do have Decky but I don't use it for anything that would affect the game. I don't use decky recorder or anything like that. Mostly just for CSS Loader

BbastiHST90 2024-02-13 github

Try protonGE 8-17 that worked flawlessly and let me finish the game.

Nnshank63 2024-02-13 github

Okay I'll try it after work thanks. I just install it like I would the latest Proton GE? I.e. using protonup-qt?

BbastiHST90 2024-02-13 github

Yes, you can choose what version you want to install. And 8-17 was published right at the time when stock proton 8.0-4 released and that works.

Nnshank63 2024-02-13 github

okay I'll give it a shot and report back in about 8 hours

Nnshank63 2024-02-13 github

Do we know if this is only happening on OLED steam decks? I have an OLED and some people above mentioned they have an OLED.

Someone on reddit said he has been playing Witcher 3 DX12 for many hours the last few days on an LCD deck and doesn't have any freezing issues in the menu.

BbastiHST90 2024-02-13 github

I mean I still had a couple crashes but no freezes. Mostly because of some depth of field effects I think. fyi

I also use OLED. No idea if LCD has the same problems.

BbastiHST90 2024-02-13 github

Ah no.. @Mohandevir wrote that he uses an LCD deck

https://github.com/ValveSoftware/Proton/issues/607#issuecomment-1927248216

MMohandevir 2024-02-13 github

Got the exact same freeze on stock lcd Deck and PC with Holoiso. Solved it
with Proton 8.4-GE.

Le mar. 13 févr. 2024, 9 h 37 a.m., nshank63 @.***> a
écrit :

Do we know if this is only happening on OLED steam decks? I have an OLED
and some people above mentioned they have an OLED.

Someone on reddit said he has been playing Witcher 3 DX12 for many hours
the last few days on an LCD deck and doesn't have any freezing issues in
the menu.

—
Reply to this email directly, view it on GitHub
https://github.com/ValveSoftware/Proton/issues/607#issuecomment-1941654258,
or unsubscribe
https://github.com/notifications/unsubscribe-auth/ABQFRY7TE5BP67WRZXF2MSLYTN3CBAVCNFSM4FRSLUZ2U5DIOJSWCZC7NNSXTN2JONZXKZKDN5WW2ZLOOQ5TCOJUGE3DKNBSGU4A
.
You are receiving this because you were mentioned.Message ID:
@.***>

Nnshank63 2024-02-13 github

thanks. so strange. Between this and the mass effect legendary edition no longer working on steam deck since the latest EA update its like the universe doesn't want me to play my favorite games lol.

I'll try the proton versions suggested later today hopefully they resolve my issue as they did yours.

Mmatheusfaustino 2024-02-14 github

I can also testify that Proton 8.4-GE is the only version (with HDR support) that works. If I use DX12 with a proton version without HDR, it also works without issue. The problem starts to happen when I use a Proton version that supports HDR.

Nnshank63 2024-02-14 github

Sorry forgot to report back yesterday

proton ge 8-17 worked for me

running dx12 for many hours no issues

Iianwhiscombe 2024-02-25 github

Another vote for 8-17. Fixed the issue.

MMohandevir 2024-03-21 github

Partially fixed with 8-4GE and 8-17GE. I get lots of crashes.
This said, I think I discovered that the initial problem is linked to the in-game fps cap. If it is set to unlimited, this is where the game freeze happen, with official Proton 8-5. Did a delete proton file and set the in-game fps cap to 60 (even if Gamescope is set to 45fps)... It's seems rock solid.

Mmoonman5 2024-03-22 github

@Mohandevir I tried to do as you wrote (official 8.0-5 proton, deleted proton files, set the in-game fps cap to 60 and Gamescope cap to 45 fps) and at first it actually worked. The game did not freeze during gwent and in the bestiary. Unfortunately, after a short while it began to freeze again, but this time after entering any menu.

MMohandevir 2024-03-23 github

Yep! Same here. Sad... I Tought I was on to something... Back to
GE-Proton-8.4, with in-game fps cap.

GE-Proton-8.17 is crashing way too often and randomly.

Le ven. 22 mars 2024, 7 h 32 p.m., moonman5 @.***> a
écrit :

@Mohandevir https://github.com/Mohandevir I tried to do as you wrote
(official 8.0-5 proton, deleted proton files, set the in-game fps cap to 60
and Gamescope cap to 45 fps) and at first it actually worked. The game did
not freeze during gwent and in the bestiary. Unfortunately, after a short
while it began to freeze again, but this time after entering any menu.

—
Reply to this email directly, view it on GitHub
https://github.com/ValveSoftware/Proton/issues/607#issuecomment-2016130735,
or unsubscribe
https://github.com/notifications/unsubscribe-auth/ABQFRY75Y6BUGKSSSP7772TYZS5RNAVCNFSM4FRSLUZ2U5DIOJSWCZC7NNSXTN2JONZXKZKDN5WW2ZLOOQ5TEMBRGYYTGMBXGM2Q
.
You are receiving this because you were mentioned.Message ID:
@.***>

Kkisak-valve maintainer 2024-03-23 · hidden on GitHub github

The Witcher 3 - Freeze in Inventory

Issue transferred from https://github.com/ValveSoftware/Proton/issues/7597.
@rariema posted on 2024-03-23T11:32:02:

Compatibility Report

  • Name of the game with compatibility issues: The Witcher 3
  • Steam AppID of the game: 292030

System Information

  • GPU: RX 6000
  • Video driver version: Mesa 24.0.3
  • Kernel version: 6.8.1-arch1-1
  • Proton version: 9.0.5 Beta

I confirm:

  • [x] that I haven't found an existing compatibility report for this game.
  • [x] that I have checked whether there are updates for my system available.

Symptoms

Game freezes in Inventory

Reproduction

Start Game, load savegame.
Open Inventory, Map, Charakter etc. Do some stuff. Game will stop responding to inputs after a while.

Possible regression, game runs fine with Proton 8.0.5

steam-292030-witcher3.log

Ssimifor 2024-03-25 github

@rariema from your logs it seems you're using dx11 mode so that's what I used, and I couldn't reproduce your issue. I tried a few sessions, one of them half an hour as you didn't specify how long a while is, but I never managed to reproduce the issue, both gamepad and m&kb remained functional through my testing.

Of possible note is the fact that you have amdvlk installed and that the launcher is using it. If you don't have it installed for any particular reason, I'd suggest uninstalling it to see if anything changes.

Rrariema 2024-03-29 github

@simifor thx for looking into it. i'm using dx12, never had this with dx11/dxvk.

after i reported this, i got the same issue on proton8. dont know why it never happend on my playthrough the weeks before.
tried much, uninstalled the 32bit amdvlk, skipped the launcher, tried x11 (normally i use plasma (6) on wayland), gnome, lts kernel, older mesa, reinstalled the game. still same issue. finally i started a gamescope session from sddm, and baam, no freeze.

turns out its an issue with the mangohud framelimiter, if its active, game freezes in inventory, normally after a few clicks of doing random stuff.

without it beeing active at game startup, the witcher 3 runs fine on proton8.0.5c an 9.0-beta16.
if the limiter is enabled later on, the issue does not occur, until you alt-tab out and in again.

so its not proton9 related, not the best report and a wrong guess :)

thx for all the work on linux gaming to everyone involved, really appreciate it.

Mmoonman5 2024-03-29 github

I can confirm what @rariema found out. Turning off gamescope framelimiter before starting game from launcher and turning it on while the game is already running does help. No freezes and everything works like it should.

Edit: After some more testing the game froze while checking the map. It looks like gamescope framelimiter must be turned off and the issue occurs only when it's turned on.

RRaduAvramescu 2024-05-04 github

I'd like to add additional information, based on this issue: https://github.com/flightlessmango/MangoHud/issues/1302#issuecomment-2094338343

With Mangohud's framelimiter, I also had the freezing issues with Fedora 40 KDE (either with Proton Experimental, or GE-Proton 9-4, didn't try others). With a fresh install of Fedora Workstation 40 (so, Gnome), the game no longer freezes.

I'm thinking maybe it has something to do with KDE. That said, you said you had these issues with Gnome @rariema? Was it a fresh install or you installed it side-by-side with, say, KDE?

Rrariema 2024-05-05 github

All my testing was done on my main installation, an EndevourOS based KDE installation with Gnome and gamescope-session-git installed alongside.

MMollyMaclachlan 2024-05-09 · hidden on GitHub github

Game is crashing for me once I pass the loading screen to load a save / start a new game. It gets to the title screen fine and I can do whatever I want there, but as soon as I load into the game it crashes; there's usually just enough time for it to fade in enough for me to see a hint of the game world and UI.

Tried multiple Proton versions; with and without mods enabled; with and without gamemode enabled; no differences on any configurations. I use the --skip-launcher launch command and have DX11 set as my fallback. Used to play the game with no issues at all.

Running Slackware Current, Ryzen 7 3800X & RX 6600.

Error log with Proton 9.0-1

Full system diagnostics according to steam

Kkisak-valve maintainer 2024-05-09 · hidden on GitHub github

Hello @MurdoMaclachlan, can you check if https://gitlab.freedesktop.org/drm/amd/-/issues/3343 is relevant to your system? A quick test would be to reboot into an older kernel and see how the game behaves.

MMollyMaclachlan 2024-05-10 · hidden on GitHub github

Ah, yeah, rolling back to 6.6.29 fixed the issue; I can run the game with no problems now. Thanks!

Ooakad 2024-05-26 github

Exactly the same crash issue happens with Nvidia v550.78.0/6.9.1 kernel.
(arn:seh:dispatch_exception EXCEPTION_ACCESS_VIOLATION exception (code=c0000005))

Interestingly enough, the game starts in the dx12 mode, and even works, but screen stutter is pretty aggressive, making it unplayable.

Zzaps166 2024-05-26 github

After recent game update it crashes with DX11+HairWorks (game issue, also on Radeon). Turn off HairWorks and try again.

DX12 is running about 2x slower than DX11 to me (game issue).

Ooakad 2024-05-27 github

Hairworks it is. Thank you.

Zzaps166 2024-06-06 github

Hairworks it is. Thank you.

Crash should be fixed by the new game update!

MMohandevir 2024-07-06 github

Issue still not solved. Can't join my fresh new log file (too big: 520mb). What should I do?

Edit: Just compressed it to .tar file (if allowed)
steam-292030.tar.gz

CCOLESLAW117 2024-09-23 github

Somehow, this game is still verified despite the fact that it crashes anytime the map, inventory, or bestiary are accessed in the DX12 version of the game. DX12 is needed for FSR, and more importantly, HDR, which is very important to playing this game on the OLED Steam Deck. Switching to the DX11 version is not a viable alternative.

Zzaps166 2024-09-23 github

Somehow, this game is still verified despite the fact that it crashes anytime the map, inventory, or bestiary are accessed in the DX12 version of the game.

I don't have Steam Deck, but map and bestiary are working normally on my PC (RX 6900 XT). The only thing is DX12 is over 2x slower than DX11 (game issue), so FSR/XeSS is must-have, but it reduces image quality.

CCOLESLAW117 2024-09-23 github

It seems to happen if the frame limiter is used in Game Mode. I use the quick menu controls for just about every game so I can tune the perfect performance/efficiency ratio

Zzaps166 2024-09-23 github

It seems to happen if the frame limiter is used in Game Mode. I use the quick menu controls for just about every game so I can tune the perfect performance/efficiency ratio

So we know it's external frame limiter issue. Without external frame limiter: are map and bestiary working correctly?

Rrariema 2024-09-23 github

Thats what i reported in May. Seems the problem persists. Just fininished the DLCs without a external framelimiter in DX12, FSR2-Q with nice performance and quality without a single freeze. (PC, RX 6600)

Did a quick test right now with Mangohud limiter, freeze in inventory (map, bestiarium...) is still there.

MMohandevir 2024-09-23 github

I could not test extensively to see if it solves the issue permenantly, but
it also happens with some system wide upscalers turned on. Linear seems to
work fine.

Yeah, I know, should be obvious. Thing is, my Gamescope settings were
configured for the DX11 version and I forgot about FSR. It began to be a
problem when Proton 8.0.5 was released.

Le lun. 23 sept. 2024, 11 h 55 a.m., Błażej Szczygieł <
@.***> a écrit :

Somehow, this game is still verified despite the fact that it crashes
anytime the map, inventory, or bestiary are accessed in the DX12 version of
the game.

I don't have Steam Deck, but map and bestiary are working normally on my
PC (RX 6900 XT). The only thing is DX12 is about 2x slower than DX11 (game
issue), so FSR/XeSS is must-have, but it reduces image quality.

—
Reply to this email directly, view it on GitHub
https://github.com/ValveSoftware/Proton/issues/607#issuecomment-2368711640,
or unsubscribe
https://github.com/notifications/unsubscribe-auth/ABQFRY5A34UXIVXZ6E2G2RTZYA2X5AVCNFSM4FRSLUZ2U5DIOJSWCZC7NNSXTN2JONZXKZKDN5WW2ZLOOQ5TEMZWHA3TCMJWGQYA
.
You are receiving this because you were mentioned.Message ID:
@.***>

Zzaps166 2024-09-23 github

Ok, I see freeze with mangohud and other external limiters.

CCOLESLAW117 2024-09-23 github

I'm using Proton GE-8-17, which does not have the freezing issue. The game works fine on the most recent official proton when the frame limiter is not used. Let me know if there is anything I can do to help resolve this issue.

Tthecowmilk 2024-10-03 github

Hi guys, I have these Witcher 3: Wild Hunt problems:

  • Game crashes and freezes during menu including: Alchemy, Map, Regular Menu at some merchants when you finished business with them, crafting menu, repair menu.

Everything Menu related breaks, crashes and freezes the game! I'm running it using Proton GE-9.14

System:

OS: Ubuntu 24.04.1 LTS
KERNEL: 6.8.0-45-generic
CPU: 12th Gen Intel Core i5-12450H
GPU: NVIDIA GeForce RTX 4050 Laptop GPU
GPU DRIVER: NVIDIA 555.58.02
RAM: 16 GB
Zzaps166 2024-10-03 github

Is it vkd3d-proton bug or another proton component bug - anybody knows? Maybe it's worth to bisect, so devs could fix it.

We already know Proton GE-8-17 works, does it mean Proton GE-8-18 doesn't work?

Ssimifor 2024-10-05 github

I've been unable to repro on my desktop or deck thus far. I see several mentions of GE proton, but has anyone gotten this issue with proton stable (Currently 9.0-3)? If so, could you share your graphics settings in case they might be a factor?

Also, it could be helpful to get some logs, if you add the following to the game's launch parameters: PROTON_LOG=1 %command% you'll get a log file in your home folder, just get the game to hang and close it, and then upload the resulting steam-292030.log file, compress it if it's too big.

CCOLESLAW117 2024-10-05 github

Replying to https://github.com/ValveSoftware/Proton/issues/607#issuecomment-2395081159

steam-292030.zip

Used the standard proton 9 from valve. let me know if you need anything else

Deck Quick Settings
Frame Limit: 45fps @ 90hz
Allow Tearing: Yes

Graphics Settings (using DX12, ver 4 .04):
Ray Tracing: Off
Anti-Aliasing: FSR 2
FSR Quality: Quality
Dynamic Resolution Scaling: Off
Sharpening: High
Screen Space Ambient Occlusion: SSAO
Screen Space Reflections: Low
Motion Blur: Off
Blur: Off
Bloom: On
Depth of Field: On
Chromatic Aberration: On
Vignetting: On
Light Shafts: On
Camera Lens Effects: On
NVIDIA HairWorks: Off
Number of Background Characters: Medium
Shadow Quality: Medium
Terrain Quality: Medium
Water Quality: Medium
Foliage Visibility Range: Medium
Grass Density: Medium
Texture Quality: High
Detail Level: Medium

Display Settings
Display Mode: Full Screen
VSync: Off
Resolution: 1280x800
Maximum FPS: Unlimited
Hardware Cursor: Off

HDR Settings
Max Monitor Brightness: 1000
Paper White: 300

CCOLESLAW117 2024-10-05 github

Replying to https://github.com/ValveSoftware/Proton/issues/607#issuecomment-2395119074

Playing on the deck using the frame limiter btw

CCrashTD 2025-02-03 github

Unfortunately the game is very unstable on DirectX 12. No matter the used proton version (9.0.4, "hotfix" or experimental), after a seemingly random period of time (can be 1 minute, can be 1 hour), the game suddenly freezes completely and has to be closed via Steam. It also seems random in what situations. Just walking around, or being in the inventory or talking to someone. This seems to be in line with other reports I have read on protondb.

Log: steam-292030.log

System Information

  • GPU: RX 7800 XT
  • CPU: Ryzen 7 3700x
  • Video driver version: Mesa 24.3.4
  • Kernel version: 6.12.10-arch1-1
  • KDE Wayland
  • Proton version: 9.0-4, Hotfix, Experimental

Using DX11 solves the problem but locks one out of the nice new next-gen graphics.

Mmad526 2025-05-13 github

Same instability for me on DirectX 12. Tested proton version 9.0.4 and 10.0-1 beta. Freezes of the game occur after a random amount of time. I noticed the likelihood of a freeze early on to be much higher when RT is enabled, but they are still occuring even when raytracing effects are all disabled.

System Information

  • GPU: AMD Radeon RX 9070 XT
  • CPU: AMD Ryzen 7 5800x
  • Video driver version: Mesa 25.1.0
  • Kernel version: 6.14.4-104.bazzite.fc42.x86_64
  • Proton version: 9.0-4, 10.0-1 beta
DDerEchteJoghurt 2025-05-13 github

I'm having the exact same instability on DirectX 12 as the other people here. Sometimes after a random amount of time the game freezes completely with audio still playing. No freezes on DX11, but I would love to play with raytracing.

System Information

GPU: NVIDIA GeForce RTX 4070 Super
CPU: AMD Ryzen 7 3700X
Video driver version: nvidia-open 570.144-5
Kernel version: 6.14.6-arch1-1
Proton version: 9.0-4, 10.0-1 beta, GE Proton 9 27
Kkisak-valve maintainer 2025-05-17 github

The Witcher 3: Wild Hunt (292030) freezes

Issue transferred from https://github.com/ValveSoftware/Proton/issues/8704.
@crazycorndog posted on 2025-05-17T00:12:48:

Compatibility Report

  • Name of the game with compatibility issues: The Witcher 3: Wild Hunt
  • Steam AppID of the game: 292030

System Information

I confirm:

  • [x] that I haven't found an existing compatibility report for this game.
  • [x] that I have checked whether there are updates for my system available.

Symptoms

Game freezes (video freezes, music keeps playing, cpu and gpu load drops) after up to about 2 minutes. Sometimes within a few seconds, sometimes within a few minutes. I can change windows to Steam and there force close the running game.

Launch options: %command% --launcher-skip

Reproduction

Start the game, load last savegame or continue play and wait up to 2 minutes or so.


@crazycorndog commented on 2025-05-17T00:14:21:

Proton log attached steam-292030.log

Ccpuccino 2025-06-24 github

Can't launch witcher 3 with default settings. (Can be bypassed by disabling red launcher)

RTX 4090, i9-13900KF
Driver: Nvidia 575.51.02
Proton version: Default
Kernel: 6.14.0-22-generic


---------------------------
REDlauncher
---------------------------
Failed to create OpenGL context for format QSurfaceFormat(version 2.0, options QFlags<QSurfaceFormat::FormatOption>(), depthBufferSize 24, redBufferSize -1, greenBufferSize -1, blueBufferSize -1, alphaBufferSize 8, stencilBufferSize 8, samples 0, swapBehavior QSurfaceFormat::DefaultSwapBehavior, swapInterval 1, colorSpace QSurfaceFormat::DefaultColorSpace, profile  QSurfaceFormat::NoProfile) .
This is most likely caused by not having the necessary graphics drivers installed.

Install a driver providing OpenGL 2.0 or higher, or, if this is not possible, make sure the ANGLE Open GL ES 2.0 emulation libraries (libEGL.dll, libGLESv2.dll and d3dcompiler_*.dll) are available in the application executable's directory or in a location listed in PATH.
---------------------------
OK   
---------------------------
Zzaps166 2025-06-24 github

Replying to https://github.com/ValveSoftware/Proton/issues/607#issuecomment-2998798378

Launcher is 32-bit, so you need 32-bit drivers.

Ccpuccino 2025-06-26 github

Getting constant freezes - every 10 minutes or so - game is unplayable.

steam-292030.log
steam-292030.log

RTX 4090, i9-13900KF
Driver: Nvidia 575.51.02
Proton version: Default
Kernel: 6.14.0-22-generic

Ssimifor 2025-06-26 github

@cpuccino A couple of questions

  1. Does the game unfreeze after freezing? or do you have to close it after it freezes?
  2. Does it freeze with both the dx11 and dx12 renderers? (when its launcher pops up you can select between the two with the cog icon)
  3. Does the game freeze if you idle ingame or do you have to actively play?
Ccpuccino 2025-06-27 github

@simifor

  1. Doesn't unfreeze - although the max I've waited for is a few minutes or so. - Alt+f4 doesn't work either. Need to close via steam.
  2. I've only played using dx12 (which I assume is the default) - the launcher doesn't work for me. Ref https://github.com/ValveSoftware/Proton/issues/607#issuecomment-2998798378
  3. I don't think i've been "idle" enough in the game to know. Only seen it freeze while doing something.
Zzaps166 2025-06-27 github

To run in DX11 mode without launcher, edit launcher-configuration.json and set fallback to DirectX 11.

Ccpuccino 2025-06-27 github

I think I'm gonna hold off on this title for now. Was mostly looking to play this on the next gen update. It looks quite different with dx11 - and from what I've heard raytracing etc aren't supported on the dx11 version. If there's any other logs needed, let me know.

Ssimifor 2025-06-27 github

@cpuccino did you install the 32 bit drivers for your card? The package name will change depending on what distro you're on, on arch based distros it'd be lib32-nvidia-utils for example.

I tried to repro on my amd desktop and nvidia laptop, around an hour each, with no luck. There are a couple of things that could be affecting the issue:

  1. Game settings, when you have the chance, please share yours.
  2. Your ingame location, could be an issue that only affects a portion of the game. Share your save file when you can.
  3. Your setup, including your driver version, my nvidia laptop is on nvidia-open 575.64, which is newer than what you're using, it's possible the issue only exists in older nvidia drivers.
Ccpuccino 2025-06-28 github
  1. Currently using Ultra without raytracing, and no hairworks.
  2. Location is around Velen. - Just randomly freezes while walking, fighting anywhere.
  3. My setup is listed above along with the logs up to the crash; 575.51.02 is the latest one currently on the graphics ppa.
Ssimifor 2025-06-30 github

@cpuccino galloped around velen for nearly an hour and killed some enemies here and there, no freeze and no oddities on my side. At this point driver issue appears to be the most likely scenario, so keep an eye for updates on your distro

Ccpuccino 2025-07-01 github

@simifor https://github.com/ValveSoftware/Proton/issues/607#issuecomment-2875050087 - lots of similar reports on both the nvidia and amd side - on this issue tracker, as well as forums so i'm a bit skeptical that it's specific to drivers.

Regardless, this seems to be a very common issue that happens on specific setups when running DX12. Let me know if there's anything in my logs that stands out and I can help debug further.

MMaselkov 2025-07-01 github

After launching the game using DX12, it plays normally for about 30 seconds or so after loading in- and then, the frametime graph becomes an almost perfect sawtooth.
Image
In-game settings don't seem to affect it. I've disabled vsync, limiter, turned the settings to lowest, turned the settings to highest.
I've tried proton experimental, latest GE, proton 9
Specs:

  • 9070 xt
  • 9800x3d
  • 64gb ram

System:

  • Linux 6.15.4-arch1-1
  • KDE Plasma 6.4.1
  • Mesa 25.2.0-devel (git-e236a731e4)

DX11 doesn't have the same issue. Anyone have any clue what could be the cause? I've seen one report of a user on reddit experiencing the same thing on a Nvidia GPU, so I don't think it's Mesa related?

It doesn't seem occur in every in-game location too. I can consistently reproduce it at the start of the game; in White Orchard, and in Corvo Bianco, for example (Among others).

Ssimifor 2025-07-03 github

I've now spent a couple of hours with this game on rdna2, rdna4 and ampere with no freezes, considering the game has been out for a long time now, are those affected by the freezes using old save files? If so, h ave you tried to see if the same issue happens on a new save file? Asked for a save file earlier but didn't see any being provided.

@Maselkov my rdna4 runs of the game have been with mesa git 3853f72f525 (slightly newer than yours) and proton 10, with the game's RT graphic preset and dynamic scaling off and vsync off, on plasma wayland. No sawtooth to be seen in the white orchard on my end.

Just to confirm, you can feel the game being stuttery, right? Simply to rule out mangohud misreporting the frametimes. If so can you give more information about your setup? Like desktop environment, or if you have multiple monitors connected. Have you tried to launch the game without background programs to see if they make a difference?

MMaselkov 2025-07-03 github

@simifor I forgot to mention, I did actually notice that enabling any degree of RT makes the sawtooth go away. .. But it's a bit of a moot point in my case, since with RT the game is unplayable in the first place due to low framerate at my resolution (3440x1440).

The save game doesn't seem to matter. I was able to figure out a way to reproduce this issue consistently, namely simply starting a new game, from the beginning, skipping through cutscenes, and idling in the Kaer Morhen chambers where you spawn in for about 16 seconds. Can record a video if needed.

Settings don't seem to matter, dynamic scaling off, or on, vsync off, so on.

As for the last part.. I am pretty sure it is a bit stuttery, particularly as I notice some more VRR flicker than usual. Although, it is hard to tell at such high fps, so I am not fully confident - but, I haven't heard of mango misreporting frame timing beforehand, so I believe it's correct.

I've shared DE info above, and yes, I am running multiple monitors, but it happens even if I only have only my single 3440x1440p 144hz monitor. Background programs have been assassinated as well, although no particular core was pushed to its limits really.

Ssimifor 2025-07-05 github

@Maselkov When I asked about desktop environment I mean if you were using wayland on xorg, but it doesn't matter anymore I could repro on plasma wayland 6.4.2. However, I get the impression that mango might be misreporting frametimes here for two reasons:

  1. Panning looks smooth, I'd expect hitches from such severe variations
  2. My monitors brightness isn't changing. My display suffers from VRR flicker, brightness changes visibly when FPS has high variations, but the brightness looks constant to my eyes in this case

https://github.com/user-attachments/assets/fe90ed51-0e76-43a1-b290-868b94ec5ba9

MMaselkov 2025-07-05 github

@simifor Oh, yeah, my bad, sorry!

Anyway, interesting findings. In general, something -feels- off to me.. but that could very well just be confirmation bias from having seen that graph. Perhaps it might be mangohud misreporting. I wonder if there's any way this could be confirmed..

Kkisak-valve maintainer 2025-07-20 github

Frame Generation makes the game stutter while showing 60fps on Mangohud with Cyberpunk and Witcher 3

Issue transferred from https://github.com/ValveSoftware/Proton/issues/8918.
@ShepardCommander100 posted on 2025-07-20T08:39:15:

Compatibility Report

  • Name of the game with compatibility issues: Witcher 3 and Cyberpunk 2077
  • Steam AppID of the game: 292030 and 1091500

System Information

  • GPU: RTX 4060 Mobile
  • Video driver version: nvidia-open-dkms 575.64.03
  • Kernel version:6.15.7-arch1-1
  • Link to full system information report as Gist:
  • Proton version: Proton 10, Proton Experimental, Proton GE 10-8 and GE 10-10

I confirm:

  • [ ] that I haven't found an existing compatibility report for this game.
  • [ ] that I have checked whether there are updates for my system available.

Symptoms

My game keeps stuttering when I turn on frame generation

Reproduction

For Cyberpunk, it happens randomly when I load a save, sometimes it happens in one load, other times it happens after 4-5 reloads.
But for Witcher 3, this happens constantly

SShepardCommander100 2025-07-20 github

Hello, thank you for the response, the reason I clubbed the cyberpunk and witcher 3 issues into 1 was that I thought that since they were from the same engine and same dev, and the issue was very similar, that there might be a common issue at play here

SSimpleHeuristics 2025-07-24 github
Kkisak-valve maintainer 2025-10-24 github

The Witcher 3 292030 - Unable to launch on CachyOS

Issue transferred from https://github.com/ValveSoftware/Proton/issues/9132.
@GrayCountertop posted on 2025-10-24T04:19:40:

Compatibility Report

  • Name of the game with compatibility issues:
  • Steam AppID of the game:292030

System Information

  • GPU: GTX 2060
  • Video driver version:
  • Kernel version:
  • Link to full system information report as Gist:
  • Proton version: Experimental, Cachyos, 9.0.4, hotfix, GE... all of them I click play, it goes blue, and then nothing.

I confirm:

  • [ yes] that I haven't found an existing compatibility report for this game.
  • [ yes] that I have checked whether there are updates for my system available.

steam-292030.log

Symptoms

Game does not launch. I've been told by others that it works for them, unsure why it doesn't work for me. Proton log attached. I click play, button turns from green to blue... I wait, nothing happens and then it goes back to "play".

I really want to play this game. It cost a lot a the time. I'm on cachyos, proton, and native steam. I tried to get it to launch without the launcher and/or with directx11 and neither worked.

Reproduction

KKlusio19 2025-11-16 github

Replying to https://github.com/ValveSoftware/Proton/issues/607#issuecomment-3025147454

Edit: I removed long part about input lag, because I just discovered that it's happening on Windows aswell, so it's not proton's fault. It's either game's problem or something else.

GPU: RX 9060 XT
@Maselkov 4 months later but here I come to tell you that I have basically exactly the same problem but also noticed one more thing!
I'm experiencing the exact same sawtooth frametime graph, after loading the game. For me it happens after 20-22 seconds. And just like in your case, it doesn't happen everywhere.

KKlusio19 2025-11-16 github

Replying to https://github.com/ValveSoftware/Proton/issues/607#issuecomment-3039511999

GPU: RX 9060 XT
vulkan-overlay shows the same crazy-looking frametime graphs.

Ssimifor 2025-11-21 github

@Klusio19 if you select the steam deck graphics preset do the sawtooth go away? If so, does that remain the case after switching out of said preset?

KKlusio19 2025-11-21 github

@Klusio19 if you select the steam deck graphics preset do the sawtooth go away? If so, does that remain the case after switching out of said preset?

@simifor After using steam deck graphics preset, sawtooth doesn't go away. It gets "smaller?" probably beacuse of higher number of FPS - check screenshots.

Uber preset:

Image

Steam deck preset:

Image
Kkisak-valve maintainer 2025-12-16 github

Linux client and running game freeze with message err:winevulkan:signaller_worker wait timed out with non-empty poll list

Issue transferred from https://github.com/ValveSoftware/Proton/issues/9316.
@rubendelmonte posted on 2025-12-16T09:36:20:

SSUE
The Witcher 3 (but other suspect as well, like Darktide) randomly freeze with repeated message "err:winevulkan:signaller_worker wait timed out with non-empty poll list"
Steam client is frozen as well. ALT+TABBing to another window does work, closing the game with System Monitor brings client back to control.
Checks in GPU usage with LACT just show a sudden drop to 0% in GPU utilization when game freezes; up to that point utilization, speed, temps and voltages are in the legit ranges, as per attachment.
No other relevant logs available.
No overlclock or undervolt applied.
Behavior has been observed with different Proton versions, both from Valve (default and 9.0.4) and GE (10.24 and 10.26).
Reproduced with and without environment variables (like "gamemoderun" or "LD_PRELOAD="" ")
Older games like Grim Dawn or lightweight ones like Demeo seems not affected.
Personal consideration: I wonder if the recently implemented NTSYNC driver has something to do with it.

HW SPECS:
MB ASUS B650 PLUS
AMD RYZEN 9 7900x
32 GB DDR 5200 KINGSTON
RADEON RX 9070XT
1x 2TB NVME
1x 2TB SDD
PSU ENERMAX REVOLUTION III 850W 80+ CERTIFIED
PC is attached to a 1200VA UPS

Note: with Darktide, the behavior has been observed also by using an HP Omen Laptop equipped with Intel Core i9-14900HX, 32GB RAM DDR5, 1TB SSD PCIe NVMe, Nvidia GeForce RTX 4070 8GB GDDR6

OS
tested both in Fedora 42, Fedora 43 and Debian 13

Currently running on Fedora 43, specs as follows:
Kernel: 6.17.11-300.fc43.x86_64
AMD firmware:
amd-gpu-firmware-20251125-1.fc43.noarch
amd-ucode-firmware-20251125-1.fc43.noarch
MESA:
mesa-libGLU-9.0.3-7.fc43.x86_64
mesa-compat-libxatracker-25.0.7-3.fc43.x86_64
mesa-filesystem-25.2.7-3.fc43.i686
mesa-filesystem-25.2.7-3.fc43.x86_64
mesa-dri-drivers-25.2.7-3.fc43.x86_64
mesa-libgbm-25.2.7-3.fc43.x86_64
mesa-libEGL-25.2.7-3.fc43.x86_64
mesa-libGL-25.2.7-3.fc43.x86_64
mesa-vulkan-drivers-25.2.7-3.fc43.x86_64
mesa-va-drivers-25.2.7-3.fc43.x86_64
mesa-va-drivers-25.2.7-3.fc43.i686
mesa-dri-drivers-25.2.7-3.fc43.i686
mesa-libgbm-25.2.7-3.fc43.i686
mesa-libEGL-25.2.7-3.fc43.i686
mesa-libGL-25.2.7-3.fc43.i686
mesa-vulkan-drivers-25.2.7-3.fc43.i686
mesa-vdpau-drivers-freeworld-25.2.7-2.fc43.x86_64
mesa-va-drivers-freeworld-25.2.7-2.fc43.x86_64
WAYLAND:
libwayland-client-1.24.0-1.fc43.x86_64
libwayland-server-1.24.0-1.fc43.x86_64
libwayland-egl-1.24.0-1.fc43.x86_64
libwayland-cursor-1.24.0-1.fc43.x86_64
libwayland-client-1.24.0-1.fc43.i686
libwayland-cursor-1.24.0-1.fc43.i686
libwayland-egl-1.24.0-1.fc43.i686
libwayland-server-1.24.0-1.fc43.i686
xorg-x11-server-Xwayland-24.1.9-1.fc43.x86_64
qt5-qtwayland-5.15.18-1.fc43.x86_64
qt6-qtwayland-6.10.1-1.fc43.x86_64
qt6-qtwayland-adwaita-decoration-6.10.1-1.fc43.x86_64
gnome-session-wayland-session-49.2-1.fc43.x86_64

STEPS
just play the mentioned games, the freeze will randomly happen

RESULTS
Game and Steam Client freeze

proton_witcher3_crash.txt
lspci.txt
lsb_release.txt
LACT-stats-20251216-091251_witcher3_crash.ods

Rrubendelmonte 2025-12-17 github

Follow-up: to exclude HW problems, I did two things: 1. install and play Shadow of the Tomb Raider on Linux, which runs natively and 2. spin-up a Windows 11 partition to play Witcher 3 with "officially supported" config. In both cases, I didn't have crashes or freezes and I observed some interesting thing: even at 100% utilization, the GPU clock never goes above 2.8/2.9GHz. As reported in the LACT charts, by playing Witcher 3 over Proton the GPU speed skyrockets at 3100+ MHz (and I can confirm 100% the same with Darktide). Seems that Proton with the mentioned games somehow overloads the GPU up to a crash.
Edit 01/04/2026: I can confirm the Witcher 3 freeze behavior also on my laptop (HP Omen Laptop equipped with Intel Core i9-14900HX, 32GB RAM DDR5, 1TB SSD PCIe NVMe, Nvidia GeForce RTX 4070 8GB GDDR6), at this point it is possible to assume that it is not hardware or configuration dependant.
Edit 01/10/2026: after several hours of testing both with my laptop and desktop, seems that forcing the game in dx11 mode solves the freeze problem. Also, the frametime results much more consistent (with almost a flat curve in mangohud) if playing in dx11 mode.

OOneShot95 2026-01-17 github

Replying to https://github.com/ValveSoftware/Proton/issues/607#issuecomment-3666217433

This issue is happening to me on a select few games on both a 4080 laptop and a 4090 desktop. Seems to only be select UE5 games.

Backrooms: Escape Together
Emissary Zero

I have tested other games that have not had any issues running that I believe is on UE5 -
Wreckfest 2
Outlast Trials
Beyond Hanwell

As mentioned, forcing DX11 seems to partially solve the issue. Although on Emissary Zero, I get fatal error crashes at the exact same moment on the laptop and desktop.

In the logs, the main error seems to be - err:winevulkan:signaller_worker wait timed out with non-empty poll list.

I'm very new to Linux and learning so I don't know why this error would appear but I'll keep testing and see if anything comes up.

Edit: When manually disabling Hardware Ray Tracing in the config before launching the game, it launches correctly.

Edit 2: Rolling back to a previous NVIDIA driver prior to the 590.48 version fixes this issue, something with the drivers causing issues.

Kkisak-valve maintainer 2026-02-01 github

The Witcher 3 Freezes While Selecting a Quest

Issue transferred from https://github.com/ValveSoftware/Proton/issues/9452.
@Gacel0l posted on 2026-01-31T23:26:21:

---GAME--
Steam AppID: 292030
PC
GPU: RTX 2060 driver 590.48.01
Computer Information:
Manufacturer: ASRock
Model: B450 Gaming K4
Type: Desktop
No touchpad detected
Processor Information:
CPU Vendor: AuthenticAMD
CPU Brand: AMD Ryzen 5 3600 6-Core Processor
CPU Family: 0x17
CPU Model: 0x71
CPU Version Number: 0x0
CPU Type: 0x0
CPU Frequency: 3599 MHz
Logical Cores: 12
Physical Cores: 6
Hyper-Threading: Supported


Problem
The game generally works well, but there are unexpected crashes and freezes. I especially noticed that when changing a quest to be completed in the game or its sub-quest, the game can often freeze completely.

While making this log, the game froze only for a moment, but it sometimes happens that during longer gameplay and changing quests, the game freezes completely. It is random, sometimes a slight freeze, sometimes nothing, and sometimes a complete freeze.

steam-292030.log

Kkisak-valve maintainer 2026-02-01 github

Hello @Gacel0l, err:steamclient:steamclient_call Access violation in ISteamInput_SteamInput006_Shutdown. looks like a line of interest from your Proton log.

JJerycho95 2026-02-11 github

Same issue here, RTX 5080 Fedora 43, KDE
err:winevulkan:signaller_worker wait timed out with non-empty poll list.

full log: https://0x0.st/PAEQ.log

JJerycho95 2026-02-22 github

I am experiencing significant performance degradation and stuttering during gameplay in The Witcher 3: Wild Hunt - Complete Edition.
Logs indicate a direct correlation between FPS drops and VKD3D-Proton's disk cache flushing and swapchain reconfiguration.
The issue most often occurs after extended gameplay, particularly when I switch between windows during the game (Discord, browser, Heroic, and other applications).

Log

8446.285:04c8:0618:info:vkd3d-proton:dxgi_vk_swap_chain_recreate_swapchain_in_present_task: Got 3 swapchain images.
8817.693:04c8:05d0:info:vkd3d-proton:vkd3d_pipeline_library_disk_thread_main: Pipeline cache marked dirty. Flush is scheduled.
8818.693:04c8:05d0:info:vkd3d-proton:vkd3d_pipeline_library_disk_thread_main: Flushing disk cache (wakeup counter since last flush = 1532). It seems like application has stopped creating new PSOs for the time being.
0588:fixme:secur32:get_cipher_algid unknown algorithm 23

0588:fixme:secur32:get_mac_algid unknown algorithm 200, cipher 23

Additionally, I am encountering the following error:
err:winevulkan:signaller_worker wait timed out with non-empty poll list

This causes the game to freeze—sometimes for several seconds, and occasionally requiring a full restart.

System Information:
GPU: NVIDIA GeForce RTX 5080 (Driver: 590.48.01)
CPU: AMD Ryzen 7 9800X3D
OS: Fedora Linux 43
Proton: GE-Proton10-32 (The results are the same as on Proton Experimental, except HDR is not functional on the Experimental branch)
Config: HDR Enabled, Fsync Enabled, Ultra+ (Max) Settings with Ray Tracing

Full log:
launch.log

My save file:
ManualSave_119500_7ea0d400_559e71c.tar.gz

JJerycho95 2026-02-27 github

Another freeze

Image

nvidia-bug-report.log.gz

Sstellaberrant 2026-03-03 github

I am having the same issue, the Witcher 3 DX12 completely freezes at random intervals, and also freezes the Linux Steam client with it. So far nothing I did has fixed it. It can run for over 1 hour before freezing or run 10 minutes without freezing, it seems kind of random. The DX11 version works fine, but it's a shame that the DX12 is so unstable because I want to use Raytracing 🥺

this appears over and over again when the freeze happens err:winevulkan:signaller_worker wait timed out with non-empty poll list.

I tried with basic Proton as well, currently GE-32 to see if it makes things better but it hasn't so far

device-info.txt
log-this-boot.txt
steam the witcher 3 log.txt
steam-292030.log

JJerycho95 2026-03-03 github

@kisak-valve We kindly request to set the fix process in motion

Nnebadon2025 2026-03-03 github

I had the game crash yesterday after 30 minutes but I had no logging enabled, I tried again today to get the proton log and i made it 1 hour and 44 minutes before it crashed.

I was using GE-Proton10-32.

steam-292030.log

system-info.txt

JJerycho95 2026-03-04 github

I’ve found a location where the game freeze very frequently. It’s a cave in Skellige where, while swimming, the game constantly freezes up completely and requires a restart. I believe this should be easy to reproduce for anyone. I’m attaching the current save file.

AutoSave_11e177_7ea10c00_4c6dd0a.tar.gz

Image
Sstellaberrant 2026-03-14 github

Replying to https://github.com/ValveSoftware/Proton/issues/607#issuecomment-3999335044

I downloaded your savefile and it did eventually freeze, but not really any faster than somewhere else. After 10 and 14 minutes, respectively. I think it's just random when it freezes tbh.

JJerycho95 2026-03-14 github

Replying to [#607 (comment)](https://github.com/ValveSoftware/Proton/issues/607#issuecomment-3999335044)

I downloaded your savefile and it did eventually freeze, but not really any faster than somewhere else. After 10 and 14 minutes, respectively. I think it's just random when it freezes tbh.

True, it’s likely due to the driver change. I’ve noticed it’s awful on 595.45.04, poor on 590.44.01, and fine on 580.142. There is an ongoing thread about this on the NVIDIA forums.

Valve should investigate, as AMD users are reporting it too.

Sstellaberrant 2026-03-14 github

I'm not even on Nvidia, i'm using a 7900xt

Sstellaberrant 2026-03-14 github

I had hoped that Mesa 26 would improve things but unfortunately it hasn't improved the freezing here

Sstellaberrant 2026-03-14 · hidden on GitHub github

I'm at the 1h mark in the cave without freezing, I added SteamDeck=1 to my launch options. I changed nothing else. Previously I couldn't even last 1/4 of that time. coincidence?

Sstellaberrant 2026-03-14 · hidden on GitHub github

scratch that, it froze after 1h27 :(

AAJuujarvi 2026-03-16 github

@stellaberrant @Jerycho95 Do you both have ray tracing turned on? I've found that ray tracing has a VRAM memory leak on Windows. The fps is really good and then it suddenly tanks as the VRAM is nearly full. After that it gets worse and worse before the game crashes to a game error window. I can trigger it after 1-2 hours of gameplay, but in some situations I can trigger it in 10-15 minutes by just standing around like in Novigrad at the square southeast of the Passiflora.

Sstellaberrant 2026-03-17 github

@stellaberrant @Jerycho95 Do you both have ray tracing turned on? I've found that ray tracing has a VRAM memory leak on Windows. The fps is really good and then it suddenly tanks as the VRAM is nearly full. After that it gets worse and worse before the game crashes to a game error window. I can trigger it after 1-2 hours of gameplay, but in some situations I can trigger it in 10-15 minutes by just standing around like in Novigrad at the square southeast of the Passiflora.

Nah this is different, it's not a crash but a full freeze with no error message. I have 20GB VRAM and at no point is it fully exhausted when playing the game, i play with mangohud overlay all the time. Usually VRAM is around 10-12gb

Aalasky17 maintainer 2026-03-17 github

Sometimes the failure behavior is not identical between Windows and Proton even if the cause is the same. It would be great to verify regardless whether or not the freeze happens only with ray tracing as that can be a hint.

Sstellaberrant 2026-03-17 · hidden on GitHub github

Sometimes the failure behavior is not identical between Windows and Proton even if the cause is the same. It would be great to verify regardless whether or not the freeze happens only with ray tracing as that can be a hint.

Understood, will try to check without RT later

Sstellaberrant 2026-03-17 github

welp it hasn't frozen without RT yet in 195 minutes of playtime

Kkristaps777 2026-04-06 github

Judging by the latest posts, the game still has crashes even in 2026?
I seem to be in the same boat, returned to the game after a couple of years, 1st time running it on linux, and experienced multiple random hangs at seemingly different intervals.
Tried Experimental vs GE, no difference, tried HairWorks off, no difference.

I gather the only viable approach is to switch to the nonRT version? No flags or env variables that can save the day?

JJerycho95 2026-04-07 github

Judging by the latest posts, the game still has crashes even in 2026? I seem to be in the same boat, returned to the game after a couple of years, 1st time running it on linux, and experienced multiple random hangs at seemingly different intervals. Tried Experimental vs GE, no difference, tried HairWorks off, no difference.

I gather the only viable approach is to switch to the nonRT version? No flags or env variables that can save the day?

For NVIDIA users, I identified two workarounds: downgrading the NVIDIA driver to the 580.xx branch, or disabling Ray Tracing.
On the latest drivers, enabling Ray Tracing is what triggers the game to freeze.

AMD users are also reporting issues, so AMD, or Proton itself probably has some fixes to make as well.

Kkristaps777 2026-04-07 github

I have an AMD card, forgot to mention.
I was really hoping to find a fix within a specific Proton version. Was investigating a similar problem with RE2 Remake, where the DX12 version would also hang, and turned out Experimental had a workaround where as GE didn't.

Sstellaberrant 2026-04-07 github

I have an AMD card, forgot to mention. I was really hoping to find a fix within a specific Proton version. Was investigating a similar problem with RE2 Remake, where the DX12 version would also hang, and turned out Experimental had a workaround where as GE didn't.

I have an AMD card also and it wasn't freezing for me without RayTracing. I think for now RT is off limits, until we get a fix in Proton 🥴

Kkristaps777 2026-04-07 github

I will continue testing with other Proton versions I still did not try. Currently, I played ~1h with RT on, no crash on 10.0-4.

Nnebadon2025 2026-04-07 github

I will continue testing with other Proton versions I still did not try. Currently, I played ~1h with RT on, no crash on 10.0-4.

I have had it take 1 hour and 45 minutes for this crash to occur, sometimes it happens much much faster though.

JJerycho95 2026-04-07 github

@kisak-valve can we ask for look into?

LLinux-Fan 2026-04-30 github

I have the same issue playing Witcher 3 on CachyOS with nVidia GeForce RTX 4090 Laptop GPU. The best solution so far is to downgrade the driver to version 580.142 orad disable Ray Tracing. The third option - using proton-cachyos - seems to be the best while using driver branches 590/595, although then crashes do occur.

44A6F6F6E61 2026-05-04 github

I have a GeForce RTX 4080 Super and am experiencing the same issue. After a while, the game just suddenly freezes, and I have to kill the process. I tried a lot of different Proton versions, including Proton cachyOS 11, but none helped. It just randomly happens after a few minutes. This only happens in Witcher 3; every other game works flawlessly.

Kkisak-valve maintainer 2026-05-05 github

The Witcher 3 (292030) under bare Xorg session

Issue transferred from https://github.com/ValveSoftware/Proton/issues/9744.
@mohansur posted on 2026-05-05T06:34:37:

Compatibility Report

  • Name of the game with compatibility issues: The Witcher 3: Wild Hunt
  • Steam AppID of the game: 292030

System Information

  • GPU: AMD Radeon RX 6750 XT
  • Video driver version: Mesa 25.2.8 (radeonsi)
  • Kernel version: 6.17.0-22-generic
  • Link to full system information report as Gist: (to be attached)
  • Proton version: Tested on Proton 6.3-8, 7.0-6, 8.0-5, 9.0-4, Proton Experimental, and Proton Hotfix

Symptoms

When running The Witcher 3 under a bare Xorg session (no window manager), input events are not correctly received by the game regardless of the Proton version used.

Proton 7.0-6, 8.0-5, 9.0-4, Experimental, Hotfix:

  • Mouse pointer events: ✅ Working
  • Mouse click events: ❌ Not received by the game
  • Keyboard events: ❌ Not received by the game

Proton 6.3-8:

  • Mouse pointer events: ❌ Not received by the game
  • Mouse click events: ✅ Working
  • Keyboard events: ✅ Working
  • Note: Input worked correctly in one isolated test run, this involved changing to fullscreen, windowed mode but could not be reproduced consistently. Behavior is effectively the same as newer versions.

In all cases, the issue is fully resolved by adding any window manager to the Xorg session like gnome-shell.

This issue has also been reproduced with other games (DOOM 2016, DOOM + DOOM II) across multiple Proton versions, suggesting it is not specific to The Witcher 3 but a general Proton/Wine input handling issue
in bare Xorg sessions without a window manager.

Reproduction

  1. Start a bare Xorg session with no window manager ( sudo /usr/lib/xorg/Xorg )
  2. Launch Steam
  3. Launch The Witcher 3 (AppID: 292030) with any Proton version
  4. Attempt mouse clicks or keyboard input in-game
  5. Observe that mouse pointer movement works (on Proton 7.0+) but clicks
    and keyboard input are not registered

Workaround: Launch the Xorg session with any window manager present ( eg. mutter )

Proton Log:

witcher3-proton-experimental.log

Ggordan-ramsey 2026-07-29 github

To run the game for me, i used the following launcher options (since i have to use game scope to get 3440x1440 resolution.
Edit: finished the entire base game (135 hours) without issues.

LD_PRELOAD="" gamescope -f -W 3440 -H 1440 -r 60 --force-grab-cursor -- %command% --launcher-skip

Image

-If you don't use --force-grab-cursor then the cursor will escape if your desktop setup has multiple monitors.
-If you don't use --launcher-skip while using --force-grab-cursor the game crashes on startup. when i added this flag "--launcher-skip" it seems that the launcher that comes with the game doesn't like the force grab cursor flag.

Tested with Proton 11.0

Hardware:
3 Monitors (Samung S34E790C, Samsung S32D850T, ASUS VE278) all connected through displayport
CPU: AMD Ryzen 7 3700x
MB: ASUS X570-l
Storage: 2x 2TB M.2 SSDs Sk hynix Platinum P41 SSD PCI-e 4.0.
RAM: 64GB DDR4 @ 3000Mhz
GPU: Nvidia RTX 3060 (Driver Version: 580.173.02 CUDA Version: 13.0)
Mouse: wireless LogiTech M170
OS: Pop! 24.04
Kernel: 7.0.11-76070011-generic

Image Image
Nnagornuk2004 2026-08-20 github

I played this game a year ago and encountered these issues:

  • The game would randomly crash. Sometimes after a few minutes, sometimes after a couple of hours. What fixed it seemingly (and somehow) was disabling GOG cloud saves. No crash afterwards. Maybe this is still an issue.
  • An issue that is still def there is freezing in menus with DX12. That can be fixed with this env var:
    VKD3D_DISABLE_EXTENSIONS=VK_KHR_present_wait

Disabling cloud saves and using that env var makes the game work great in DX12, at least without RT. I haven't tested that.

Aarthuralvesdev 2026-09-02 github

I am seeing what appears to be the same DX12 freeze with an external frame limiter on a current NVIDIA/KDE/Wayland setup.

The rendered image completely freezes, while the music continues playing and witcher3.exe remains alive. After alt-tabbing away from the frozen game, nvidia-smi pmon showed no GPU activity from witcher3.exe across three
consecutive samples. The process was sleeping in ntsync_schedule. I found no NVIDIA Xid/NVRM error in the kernel log.

System information

  • Distribution: CachyOS
  • Kernel: 7.2.2-1-cachyos
  • Desktop: KDE Plasma 6.7.4
  • Session: Wayland
  • GPU: NVIDIA GeForce RTX 4070
  • NVIDIA driver: 610.57.04
  • MangoHud: 0.8.4-1+
  • Proton: Experimental 11.0-20260826
  • Game renderer: DX12 / VKD3D-Proton
  • Primary display: 1920x1080 at 143.98 Hz
  • Secondary display: 1920x1080 at 60 Hz
  • VRR: disabled

Relevant MangoHud configuration:

fps_limit_method=early
fps_limit=0,60
toggle_fps_limit=Control_L+z
show_fps_limit

I cannot yet determine whether fps_limit_method=early specifically matters, or whether any external frame limiter triggers the problem. The symptom matches earlier reports where the DX12 renderer freezes while audio and the process
remain active.

I also noticed the recent workaround suggesting:

VKD3D_DISABLE_EXTENSIONS=VK_KHR_present_wait

Has this been confirmed as the recommended workaround when using an external frame limiter?

Kkisak-valve maintainer 2026-09-29 github

The Witcher 3: Wild Hunt — Remastered (292030) — DLSS/RR/FG unavailable: slInit() is never called and no Streamline plugin is loaded

Issue transferred from https://github.com/ValveSoftware/Proton/issues/10194.
@carlosmsalado posted on 2026-09-29T12:34:53:

Name of the game with the issue

The Witcher 3: Wild Hunt — Remastered (Steam AppID 292030), build 25575366. The appmanifest_292030.acf app name is The Witcher 3: Wild Hunt — Remastered, and launcher-configuration.json ships a single entry (DirectX 12, fallback DirectX 12) with "editions": [{"name": "remasteredEdition"}] — i.e. this edition is DX12-only (there is no bin/x64 DX11 executable on disk).

Proton version

Reproduced identically with two Proton builds, so this does not look like a Proton-version-specific regression:

  • Proton Experimental — experimental-11.0-20260924 (11.0-100), dxvk-nvapi reported at runtime as 14.0.0
  • GE-Proton11-7 — dxvk-nvapi reported at runtime as v0.9.2

Link to full Proton log

Gist: https://gist.github.com/carlosmsalado/785b4d391034f130b93fb32bea873a77

Raw: https://gist.githubusercontent.com/carlosmsalado/785b4d391034f130b93fb32bea873a77/raw/462ad4e2ce85574b3736dc3cb4223cfbbc607521/steam-292030.log

Captured with PROTON_LOG=1 (2.5 MB / 25,683 lines). Exact launch options used for the capture:

SL_ENABLE_CONSOLE_LOGGING=1 SL_LOG_LEVEL=10 SL_LOG_PATH=C:/sldbg PROTON_LOG=1 VKD3D_CONFIG=extra_rtas_sync %command% --launcher-skip

System Information

  • Distro: Omarchy (Arch-based, rolling)
  • Kernel: 7.2.5-1-cachyos
  • GPU: NVIDIA GeForce RTX 3080 (GA102, 10 GB) — PCI 01:00.0
  • Driver: 610.57.04 (proprietary, nvidia-open-dkms)
  • Integrated GPU (AMD Granite Ridge) disabled in BIOS — the system is single-GPU at test time (/dev/dri exposes only the NVIDIA card, amdgpu not loaded). This is therefore not a hybrid-GPU issue.
  • Motherboard: MSI MAG X870 Tomahawk WIFI (MS-7E51), BIOS 1.A82
  • Display: DisplayPort, 3440x1440 @ 240 Hz (Alienware AW3425DW)
  • Compositor: Hyprland (Wayland)

Steps to reproduce

  1. Launch The Witcher 3: Wild Hunt — Remastered (DX12 renderer) from Steam on an NVIDIA GPU under Proton.
  2. Open Options → Video → Anti-aliasing (upscaler list).
  3. DLSS, DLSS Ray Reconstruction and DLSS Frame Generation are all reported as not available / "not compatible".

No crash. The game renders fine (including ray tracing); only the NVIDIA upscalers are unavailable. The problem is permanent across restarts and after verifying game files.

Expected

DLSS Super Resolution selectable, as on Windows. RTX 3080 is Ampere, so DLSS-SR and DLSS-RR are hardware-supported.

Actual

Every DLSS-based upscaler is permanently reported as unsupported.

Evidence

(1) GPU and NVAPI are correct — this is not a GPU-detection problem:

info:nvapi64:DXVK-NVAPI v0.9.2 NVAPI gcc 14.0.0 x86_64 plain (witcher3.exe)
info:nvapi64:<-NvAPI_Initialize: OK
info:nvapi64:NvAPI Device: NVIDIA GeForce RTX 3080 10GB (610.57.4)
info:nvapi64:<-NvAPI_SYS_GetDriverAndBranchVersion: OK

(2) Only sl.interposer.dll is loaded. NONE of the Streamline plugin DLLs are ever loaded, although they exist on disk next to the interposer (The Witcher 3/bin/x64_dx12/) and are valid x86-64 PE DLLs:

Loaded L"S:\\\steamapps\\\\common\\\\The Witcher 3\\\\bin\\\\x64_dx12\\\\sl.interposer.dll"

A grep over the complete +loaddll stream for the plugin DLLs returns zero hits:

grep -c 'Loaded L".*sl\.\\(common\\|dlss\\|reflex\\|pcl\\)' steam-292030.log   ->   0

Present on disk but never loaded: sl.common.dll, sl.dlss.dll, sl.dlss_d.dll, sl.dlss_g.dll, sl.reflex.dll, sl.pcl.dll.

(3) Streamline reports the failure itself:

[streamline][error]sl.cpp:168[slValidateState] SL not initialized or no plugins found - please make sure to include all required plugins including sl.common

(4) slInit() is never called by the game. Proven by enabling Streamline's own environment-variable logging overrides (SL_ENABLE_CONSOLE_LOGGING=1, SL_LOG_LEVEL=10, SL_LOG_PATH=C:/sldbg). Streamline emits SL_LOG_HINT("Overriding logging settings from environment variables") from inside slInit(). That message never appears in the log, and C:\sldbg\\sl.log is never created. The slInit() warnings ("Seems like some DX/VK APIs were invoked before slInit()!!!" and "slInit must be called before any DXGI/D3D12/D3D11/Vulkan APIs are invoked") never appear either.

The game creates the D3D12 device at ~0.5 s and then queries Streamline feature support without ever having initialized Streamline, so it receives eErrorNotInitialized for every feature and greys out all DLSS options.

Analysis

From the upstream NVIDIA Streamline SDK source:

  • slValidateState() (source/core/sl.api/sl.cpp) returns eErrorNotInitialized when plugin_manager::arePluginsLoaded() is false.
  • arePluginsLoaded() is s_status == ePluginsLoaded || s_status == ePluginsInitialized (source/core/sl.plugin-manager/pluginManager.cpp), and s_status is set to ePluginsLoaded inside loadPlugins() — very early, before plugin enumeration. So if loadPlugins() had run even once, the validation would already pass. The observed failure means loadPlugins() never ran.
  • loadPlugins() is called from exactly one place: slInit() (return manager->loadPlugins();), and only when interposer::isEnabled() is true.
  • Conclusion: the game never calls slInit() in this environment.

The binary does use the Streamline feature-query API (GetIsDLSSSupported, GetDLSSGSupported, GetDLSSRRSupported are visible in witcher3.exe), so the integration is present — it simply never gets initialized.

One plausible trigger for the game's internal "DLSS not available" gate: the game issues four NVAPI queries whose entry points dxvk-nvapi does not implement, and gets NULL back:

info:nvapi64:nvapi_QueryInterface (0xad298d3f): Unknown function ID
info:nvapi64:nvapi_QueryInterface (0x33c7358c): Unknown function ID
info:nvapi64:nvapi_QueryInterface (0x593e8644): Unknown function ID
info:nvapi64:nvapi_QueryInterface (0xd7c61344): Unknown function ID

It is likely the game treats at least one of these as a hard capability check and disables the entire DLSS/Streamline path rather than falling back. If these entry points can be identified and implemented (or stubbed), that would probably resolve the issue.

Additional information

  • The game ships DLSS 4.x (nvngx_dlss.dll carries the NGX rel_310_9 / juicy_penguin weights), i.e. this is the August 2026 "Remastered" release that added Path Tracing, DLSS 4.5, FSR 4 and XeSS 2. The integration may simply be newer than what Wine has been validated against.
  • Driver-side NGX is present and correct: /usr/lib/libnvidia-ngx.so.610.57.04, /usr/lib/nvidia/wine/{_nvngx,nvngx,nvngx_dlssg}.dll, and the prefix has _nvngx.dll + nvngx.dll in system32.
  • Non-NVIDIA upscalers do work: the log shows amd_fidelityfx_upscaler_dx12.dll, amd_fidelityfx_loader_dx12.dll, libxess.dll and libxess_fg.dll all loading successfully. FSR / XeSS are the practical workaround meanwhile.
  • Ray tracing itself is functional with VKD3D_CONFIG=extra_rtas_sync on driver 610 (per https://forums.developer.nvidia.com/t/solution-for-witcher-3-with-rt-under-610-drivers/378760).
  • sl.interposer.json cannot be used as an override here: the string table of sl.interposer.dll shows the JSON-config code path is compiled out (production build), so enableInterposer / pathToPlugins cannot be set externally.
DDonkk0 2026-09-29 github

Same DLSS/RR/FG unavailability as the transferred #10194 report, also on an RTX 5090 with driver 615.71.09 and on today's bleeding-edge (experimental-bleeding-edge-11.0-442746-20260929).

Two things in the logs that the report above does not mention:

1. Wine never answers D3DKMTQueryAdapterInfo type 31 (KMTQAITYPE_PHYSICALADAPTERDEVICEIDS), and the game retries it eight times at a fixed interval. This happens on the main thread right after NvAPI_Initialize, before the game falls back to XeSS:

fixme:d3dkmt:NtGdiDdDDIQueryAdapterInfo type 31 not handled.   (x8, ~21 ms apart)

The same eight-try loop is in the #10194 log (RTX 3080, driver 610.57), there ~33 ms apart. A fixed-count retry right before an NVIDIA-only feature is dropped suggests the game cannot confirm the physical adapter's PCI vendor/device IDs.

2. NvAPI_Disp_GetOutputMode is not implemented in DXVK-NVAPI. Both logs show nvapi_QueryInterface (NvAPI_Disp_GetOutputMode): Not implemented method in the same sequence.

Ggabrielmaialva33 2026-09-29 github
Root cause for the DLSS/RR/FG unavailability in 5.0 ([#10194](/issue/ValveSoftware/Proton/10194) above): the game detects Wine and skips Streamline on purpose.

Static analysis of bin/x64_dx12/witcher3.exe 5.0.0.1041720 (sha256 c272b2c2e61f…), confirmed by a local test:

  1. Streamline is skipped when Wine is detected. Early in startup (CActivateState, RVA 0x1b712d0) the game does GetProcAddress(GetModuleHandleW(L"ntdll.dll"), "wine_get_version"); if that returns non-NULL it jumps over the creation of its Streamline manager, the object whose constructor calls slInit (branch at RVA 0x1b713cb). That is exactly why slInit is never called and only sl.interposer.dll gets loaded. It is not the unknown NVAPI function IDs, and not D3DKMTQueryAdapterInfo type 31: that query comes from the NVAPI static library linked into the exe and only affects its own retry logic.

  2. A second, cached Wine flag gates more features. In GPU init (RVA 0x1ee3090) the same lookup sets a byte (RVA 0x4077e7a) meaning "Wine, or Steam hardware" (ISteamUtils::IsRunningOnSteamHardware). With it set, the game:

    • forces DLSS-G "unsupported" before asking Streamline (RVA 0x1b76399), so FG stays greyed out even once Streamline runs;
    • forces the RT-supported and DXR 1.1 capability bytes to 0 (RVA 0x1ecd650), which greys out every ray tracing option;
    • skips registering the path-tracing ray-gen shaders (RVA 0x1ee4ba3);
    • reports FSR frame generation unsupported (RVA 0x2270096);
    • skips requesting 256-byte buffer alignment together with D3D12_RESOURCE_FLAG_USE_TIGHT_ALIGNMENT (RVA 0x1ee13f0).

Local verification (RTX 4090, driver 615.71.09, GE-Proton11-7, testing only, by neutralising those branches in a copy of the exe):

  • With (1) bypassed, Streamline initialises, loads every plugin, and DLSS SR / RR work.
  • Bypassing the DLSS-G gate in (2) as well makes DLSS Frame Generation work: interpolation state changed from disabled to enabled.
  • Clearing the whole flag in (2) exposes RT and path tracing, but also enables the tight-alignment request. vkd3d-proton then rejects the buffers (Invalid alignment 256 for buffer resource. Must be 0 or D3D12_DEFAULT_RESOURCE_PLACEMENT_ALIGNMENT, then GetResourceAllocationInfo3: Invalid resource desc), and the game crashes within seconds. That is probably one of the reasons CDPR tied these features to Wine detection.
  • With alignment left as under Wine, RT hits the known NVIDIA 590+ hang (vkd3d-proton #3226). With that issue's prebuild-size ×8 workaround, plus a larger game-side RT scratch buffer, it ran stable for 20+ minutes with RT, path tracing, DLSS and DLSS-G (details in #3226).

For players who want DLSS now, until a proper fix: a small patch script that checks the exe hash, keeps a backup and applies only the DLSS/DLSS-G changes by default (RT behind an explicit flag, with the caveats above): https://gist.github.com/gabrielmaialva33/33ebb2542f0513d55100b22aa2149ff5

Disclosure: my tests ran with OptiScaler loaded as dxgi.dll (my fork with DLSS 5 Neural Rendering, https://github.com/gabrielmaialva33/OptiScaler_DLSSNR), running NR on the game's DLSS output before frame generation. The Wine check is independent of it: with OptiScaler removed (A/B), DLSS was greyed out exactly the same way.

So a Proton-side fix would need the game to stop seeing wine_get_version, at least for the Streamline check (1) and the DLSS-G gate. Exposing RT under Wine additionally depends on the tight-alignment behaviour and on #3226.

SSaancreed 2026-09-29 github

@gabrielmaialva33 Fwiw, that tight alignment thingy is likely to have been fixed by https://github.com/HansKristian-Work/vkd3d-proton/pull/3308. Likely too new of a change to have made its way to any third party fork of Proton.

Ggabrielmaialva33 2026-09-29 github

@Saancreed Oh nice, thanks! That landed after the vkd3d commit GE-Proton11-7 ships, so that explains it. I'll retest on master and drop that part of the workaround if it holds.

Ggabrielmaialva33 2026-09-29 github

Confirmed, thanks! On vkd3d-proton master (4d06d745, includes #3308) the game's tight-aligned buffer request goes through fine, so that part of the workaround is only needed on older vkd3d builds like the one in GE-Proton11-7.

Kkisak-valve maintainer 2026-09-29 github

The Witcher 3 - The Wild Hunt - Remastered

Issue transferred from https://github.com/ValveSoftware/Proton/issues/10196.
@NotYourAlejandro posted on 2026-09-29T15:20:41:

Compatibility Report

  • Name of the game with compatibility issues: The Witcher 3 - The Wild Hunt - Remastered
  • Steam AppID of the game: 292030

System Information

I confirm:

  • [x] that I have checked whether there are updates for my system available.
  • [x] that I haven't found an existing compatibility report for this game.

Symptoms

NVIDIA features seem to be unavailable, preventing usage of:

  • Hairworks
  • Ray Tracing
  • DLSS
  • Frame Gen

What's also strange is that Geralt's hair doesn't seem to render, leaving him bald like Majordomo here:

Image

His baldness persists within the inventory menu:

Image

Hairwork settings are available, but it doesn't make a difference whether it's on or off:

Image

Plus ray tracing and DLSS settings are unavailable. The toast in the bottom-left of this image appears when pressing right to switch the Anti-aliasing option to DLSS:

Image

I've not had luck with env vars, PROTON_ENABLE_WAYLAND=1 and PROTON_ENABLE_NVAPI=1 didn't help unfortunately.

Reproduction

  1. Launch the game
  2. Navigate to Settings -> Video -> Graphics
  3. Turn off ray tracing if not already off
  4. Turn on ray tracing
  5. Do the same with other NVIDIA featured settings
Jjustenriquez19 2026-09-29 github

Update: my guess below turned out to be wrong: the unknown GUID is not what blocks FSR FG. See follow-up.

The Witcher 3: Wild Hunt — Remastered (v5.00b): AMD FSR 4 / FSR Frame Generation unavailable on RDNA4, game falls back to FSR 2

System

  • GPU: AMD Radeon RX 9070 XT (RDNA4, gfx1201), RADV
  • Mesa 26.2.3
  • Kernel 7.2.6 (CachyOS)
  • Proton: experimental-11.0-20260924
  • Game: v5.00b, DX12 executable (bin/x64_dx12/witcher3.exe), clean install (files verified through Steam)

Issue
Patch 5.00 adds FSR 4, FSR Frame Generation and Anti-Lag 2 for RX 7000 and newer. On older GPUs the game uses FSR 2 instead. Under Proton the game takes the FSR 2 path on an RX 9070 XT:

  • The graphics menu shows "AMD FSR 2.0 latency reduction" instead of the FSR 4 variant.
  • "AMD FSR Frame Generation" is greyed out, even with "AMD FSR Upscaling" selected as the anti-aliasing mode.

What the log shows (PROTON_LOG=1)

  • vkd3d-proton loads amdxc64.dll without errors (d3d12_device_init_vendor_hacks: Loaded amdxc64.dll successfully.).
  • The game loads amd_fidelityfx_loader_dx12.dll and amd_fidelityfx_upscaler_dx12.dll (native).
  • Right after the upscaler DLL loads, the game queries an interface that Proton's amdxc64 does not implement:
    fixme:amdxc:AmdExtStub1_QueryInterface2 unknown guid {f714e11a-b54e-4e0f-abc5-df58b18133d1}
    
    (repeated 4 times)
  • amd_fidelityfx_framegeneration_dx12.dll and amdxcffx64.dll are never loaded after that.

My guess is that FidelityFX checks for FSR 4 support through this interface. When the query fails, it treats the GPU as unsupported and falls back to FSR 2.

Already tried (no change)

  • WINEDLLOVERRIDES="amd_ags_x64=b;atiadlxx=b". The builtin AGS loads (agsInitialize is called) and atiadlxx is enabled, but the unknown GUID query still happens and the menu still says FSR 2.0.

Full log available on request.

Aarthuralvesdev 2026-09-29 github

Replying to https://github.com/ValveSoftware/Proton/issues/607#issuecomment-5893280981

same here, not works from nvidia, the game runs but with no nvidia features

Nnebadon2025 2026-09-29 github

Just a heads up, but following this guide and compiling latest VKD3D with the patch provided and flipping some bits in the exe with included python script you can get DLSS and RT working on Nvidia gpu. I haven't tried with AMD yet, other than my steamdeck which was fine.

https://gist.github.com/gabrielmaialva33/33ebb2542f0513d55100b22aa2149ff5

Image Image
Dd1g1talpump 2026-09-29 github

AMD data point for the RT part of the gist workaround: on RDNA3 (RX 7900 XTX, RADV, Mesa 26.2.3 and 26.3.0-devel),
clearing the Wine flag with the gist's --rt patches makes the game crash during loading. RADV crashes inside
vkCreateRayTracingPipelinesKHR when vkd3d-proton creates the DXR state objects (0xc0000005 in libvulkan_radeon.so,
exit code 3). This happens with Proton Hotfix, with or without OptiScaler, and with or without Fossilize.
The unpatched game runs fine.

Reported to Mesa:

Side notes for other AMD users:

  • HideWineExports (supported in proton-cachyos) is not a substitute for the gist patch. It also switches the game to
    the non-Wine buffer alignment, and the July 2026 proton-cachyos build crashes anyway because its lsteamclient lacks
    SteamUtils011, which this game needs. Use Proton Hotfix or Experimental.
  • The two DLSS/Streamline patches in the gist are useless on AMD, and with OptiScaler/fakenvapi they made the game exit
    shortly after startup here.
PPatrickSzela 2026-09-29 github

On my machine (5800X3D, RX 6750XT) with Proton Experimental/Wineland, the game performance was pretty bad. After some testing I've found that:

  • signing in with CD Project Red account causes noticeable and consistent stuttering paired with a small performance loss, even in the main menu (see attached frametime graph below); signing out fixed that issue immediately
Image
  • Screen Space Reflections on High reduce game performance dramatically - on highest possible settings (with RT & FG off), in Kaer Morhen, right after gaining control over Geralt, in the same spot and camera rotation I was getting ~76 FPS with SSR High, and ~115 FPS with SSR Low. No other setting was causing such drastic performance loss, and according to Google similar problem existed in older versions of the game

I'm not sure if these are Proton exclusive problems, or whether they're limited to my setup, so just in case I'm posting my findings for anyone suffering from similar issues.

Kkubakubakuba 2026-09-30 github

Replying to https://github.com/ValveSoftware/Proton/issues/607#issuecomment-5899876406

Thanks for this, signing out of the CDPR account practically eliminates all of the stuttering that appeared in the remaster. Dunno what is happening in the background that is causing this horrible performance - just by singing into an account.

Ooakad 2026-09-30 github

@kubakubakuba

CDPR's network client is calling home from the main rendering thread? :-D

BBlisto91 2026-09-30 github

Thank you for reporting all the regressions everyone. They are being looked at.

NNovaMoon 2026-09-30 github

Ryzen 5700x3d / RTX 3080
Proton 11 and Proton Experimental tested

Game is unplayable for me. i have constant stutters every second. the Frametime Graph looks like a very fast Heartbeat

PPatola 2026-09-30 github

I for one have a very odd issue, even after reinstalling the game and trying it with both Proton 11, Proton Experimental and Proton Experimental bleeding-edge: the launcher starts but when it tries to run witcher3.exe, this process seemingly crashes and tries to write a dump file but gets stuck on it. Same thing happens if I launch it directly by using --launcher-skip. But only on my desktop (with an AMD GPU, RX 7900 XTX); in my Legion Go 2 with CachyOS handheld and my Steam Deck, both with AMD GPUs, the game starts normally (via Proton Experimental auto-selected by steam).

BBlisto91 2026-09-30 github

@Patola Hi there. Post a Proton log from your desktop running the game with Bleeding Edge.

PPatola 2026-09-30 github

Sure, @Blisto91. System information here: https://gist.github.com/Patola/d4ad6745061c3dab7d221f12a76b88aa
summarizing: 7800X3D 64GB with RX 7900 XTX, Archlinux current, mesa 26.2.3, kernel 7.2.8-273-tkg-eevdf
Proton log is attached. I killed the game after a few minutes stuck.

steam-292030.log

BBeanzy 2026-09-30 github

@Patola After uninstalling and before reinstalling, did you verify that there was no folders/files left over for the game in .../steamapps/common?

I heard of other people not being able to boot the game up after it upgraded, because there were some residual uneeded files/folders left over from the non-remastered version that were causing issues on startup. Steam will not autoremove files from a game folder during uninstallation if those files aren't considered part of the game - it's a common thing for mod files/etc.

Ccarloslomas0 2026-09-30 github

Replying to https://github.com/ValveSoftware/Proton/issues/607#issuecomment-5893824175

Same here, but using RDNA3:

System

  • GPU: AMD Radeon 76000M XT (RADV NAVI33)
  • Driver: AMD RADV 26.1.99
  • Kernel 7.2.7 (SteamOS 3.9.2)
  • Proton Experimental
  • Game: v5.00b
Jjustenriquez19 2026-09-30 github

Follow-up to my comment above (RX 9070 XT / RDNA4, RADV, Mesa 26.2.3, Proton experimental-11.0-20260924, game 5.0.0.1041720).

FSR Frame Generation works on AMD with the Wine-flag patch plus VKD3D_CONFIG=nodxr

I applied only the two "Wine flag" patches from the gist (0x1EE2A53 and 0x1EE09F7), without the DLSS/Streamline ones:

  • Patch alone: the game crashes at the start of loading, in the same place @d1g1talpump reported on RDNA3: err:vulkan:vkCreateRayTracingPipelinesKHR Exception 0xc0000005 in Unix call (libvulkan_radeon.so + 0x1573c8). It still crashes with every RT option set to false in dx12user.settings (EnableRT, RTGIPreset, EnableRtRadiance, RTAOEnabled). The game seems to create the DXR state objects whenever the device reports DXR support, whatever the settings say.
  • Patch + VKD3D_CONFIG=nodxr: the game loads and "AMD FSR Frame Generation" can be enabled. amd_fidelityfx_framegeneration_dx12.dll gets loaded, which never happened on the unpatched exe. In a CPU-heavy spot in Novigrad it went from 81 to 128 FPS. RT is unavailable, of course.

Correction to my earlier guess: the unknown amdxc GUID {f714e11a-b54e-4e0f-abc5-df58b18133d1} is not what blocks FSR FG. The Wine flag is. The GUID query still happens with the patch applied, and amdxcffx64.dll still never loads. So the FSR 4 upscaler path looks like a separate, still unresolved problem.

BBlisto91 2026-09-30 github

The RT crashes should be fixed on Bleeding Edge.

Pperzeroo 2026-09-30 github

After dying and loading up a save, the game takes a few minutes to load (2-3mins), it hangs at the first eighth of the loading bar for a while, after that part it just loads up fine. I did some troubleshooting with an LLM and it seems to be an issue with the render thread. After a while I accidentally found a workaround, where simply by switching my workspace for a few seconds (the games rendering gets paused), it gets past that point and loads quicker, unfortunately it is still slower than v4.0.4.
I have already tried deleting shader cache, protonge, proton hotfix and reinstalling the game but nothing worked. Tomorrow I will test with a new save in case mine somehow got corrupted (the game save file was transferred from v4.0.4 to v5)

GPU: AMD AMD Radeon RX 9070 XT
CPU: AMD Ryzen 7 7800X3D
Driver Version: 4.6 (Compatibility Profile) Mesa 26.2.3-arch3.2
Kernel Version: 7.2.8-1-cachyos

PPatola 2026-09-30 github

@Patola After uninstalling and before reinstalling, did you verify that there was no folders/files left over for the game in .../steamapps/common?

Yes, there were some leftover files that I removed after uninstalling. Only then I installed again.

VValmar33 2026-10-01 github

Is there a way to hide the Wine version export on Proton Experimental?

PPatola 2026-10-01 github

Is there a way to hide the Wine version export on Proton Experimental?

I think it will be difficult to do, because the usual way windows applications do this is checking ntdll or some other component that differs substantially from Windows to Wine for wine-related functions. By doing a cursory grep on the game's files:

[4:56] [9136] [patola@risadinha The Witcher 3]% for a in ./REDprelauncher.exe ./bin/x64_dx12/sl.dlss_g.dll ./bin/x64_dx12/nvngx_dlssg.dll ./bin/x64_dx12/Galaxy64.dll ./bin/x64_dx12/sl.dlss.dll ./bin/x64_dx12/REDGalaxy64.dll ./bin/x64_dx12/GFSDK_Aftermath_Lib.x64.dll ./bin/x64_dx12/sl.pcl.dll ./bin/x64_dx12/libcurl.dll ./bin/x64_dx12/nvngx_dlss.dll ./bin/x64_dx12/witcher3.exe ./bin/x64_dx12/sl.common.dll ./bin/x64_dx12/sl.reflex.dll ./bin/x64_dx12/sl.interposer.dll ./bin/x64_dx12/sl.dlss_d.dll ./bin/x64_dx12/nvngx_dlssd.dll
do
strings "$a" | grep wine_ | xargs -i echo "${a}: {}"
done
xargs: unmatched single quote; by default quotes are special to xargs unless you use the -0 option
./bin/x64_dx12/sl.dlss_g.dll: wine_get_version
./bin/x64_dx12/nvngx_dlssg.dll: wine_get_version
./bin/x64_dx12/Galaxy64.dll: wine_get_version
./bin/x64_dx12/sl.dlss.dll: wine_get_version
./bin/x64_dx12/REDGalaxy64.dll: wine_get_version
./bin/x64_dx12/GFSDK_Aftermath_Lib.x64.dll: wine_get_version
./bin/x64_dx12/sl.pcl.dll: wine_get_version
./bin/x64_dx12/libcurl.dll: wine_get_version
./bin/x64_dx12/nvngx_dlss.dll: wine_get_version
./bin/x64_dx12/witcher3.exe: wine_get_version
./bin/x64_dx12/witcher3.exe: wine_get_version
./bin/x64_dx12/sl.common.dll: wine_get_version
./bin/x64_dx12/sl.reflex.dll: wine_get_version
./bin/x64_dx12/sl.interposer.dll: wine_get_version
./bin/x64_dx12/sl.dlss_d.dll: wine_get_version
./bin/x64_dx12/nvngx_dlssd.dll: wine_get_version
[4:57] [9137] [patola@risadinha The Witcher 3]% 

maybe sed'ing these wine_get_version to something that does not exist like pine_get_version would make it think it's running under Windows. But I haven't attempted it.

Dd1g1talpump 2026-10-01 github

Confirming RT and path tracing work on AMD with Bleeding Edge plus the exe patch.

System: RX 7900 XTX, CachyOS, kernel 7.2.8, Mesa/RADV 26.2.3, Proton experimental-bleeding-edge-11.0-443694-20261001, game 5.0.0.1041720 (Steam)

You need two things:

1. Switch Proton Experimental to Bleeding Edge. It has the vkd3d-proton/dxil-spirv fixes for the vkCreateRayTracingPipelinesKHR crash on RADV.

  • Steam Library → enable "Tools" in the filter → right-click Proton Experimental → Properties → Betas → select bleeding-edge.
  • Do this on the Proton Experimental tool, not on the game. Entering "bleeding-edge" as a code in the game's Betas tab gives "Beta access code invalid".
  • Set the game's Compatibility to Proton Experimental.

2. Patch the exe so the game doesn't hide RT under Wine. (@Valmar33, this is the answer to hiding the Wine version export.) The game calls GetProcAddress(ntdll, "wine_get_version") and greys out all RT/PT options when it finds it. Bleeding Edge does not change this. Use @gabrielmaialva33's script: https://gist.github.com/gabrielmaialva33/33ebb2542f0513d55100b22aa2149ff5

Close the game, then:

python3 w3_proton_patch.py --rt-only --dry-run   # check state, changes nothing
python3 w3_proton_patch.py --rt-only             # AMD/Intel: RT, PT, FSR FG
python3 w3_proton_patch.py --restore             # undo
  • AMD/Intel should use --rt-only. The default mode and --rt add NVIDIA-only DLSS patches.
  • The script only accepts exe 5.0.0.1041720 (sha256 checked), backs up the original, and "Verify integrity of game files" also reverts it.
  • A game update will replace the exe and remove the patch. The offsets will change too, so you'll need an updated script for a new build.
VValmar33 2026-10-01 github

@d1g1talpump

That works, thank you! Someone needs to let CDPR know that the RT stuff is fixed now, so they can remove the nonsensical check.

VValmar33 2026-10-01 github

And the patch is broken again with the new update. Sigh.

Ccarloslomas0 2026-10-01 github

Having issues to limit the FPS I tried to use Steam Game Mode Performance Overlay, the in-game option and even MangoHud. None work correctly. The one that "limits it" is the performance Overlay but if you set it at 60 it will limit your FPS to 50 FPS.

System

GPU: AMD Radeon 76000M XT (RADV NAVI33)
Driver: AMD RADV 26.1.99
Kernel 7.2.7 (SteamOS 3.9.2)
Proton Experimental
Game: v5.00b

Aarthuralvesdev 2026-10-01 github

With the new patch, the game doesnt show ray tracing or path tracing options anymore, neither dlss

PPatola 2026-10-01 github

The RT crashes should be fixed on Bleeding Edge.

Hi. Just FYI, the issue I was having particularly was simply due to missing directory ~/.local/share/Steam/steam/games. Recreating it allowed the steam client to update to a version that supports Steamutils011 required by the remaster, I even had my controller's firmware updated.

Dd1g1talpump 2026-10-01 github

Just vibed a new patch on this fork (https://gist.github.com/d1g1talpump/ada48ae4084bed65ba3a37f5de402a32) of @ @gabrielmaialva33 original gist

The 2026-10-01 game update (exe 5.0.0.1044392, Steam build 25646871) moved the patch offsets, so the original script now fails with unexpected size 90674640.

Confirmed working: RT and path tracing on AMD with --rt-only. RX 7900 XTX, CachyOS, Mesa/RADV 26.2.3, Proton Experimental bleeding-edge. The setup is otherwise the same as in my earlier comment.

What changed: the Wine check itself is the same (GetModuleHandleW(L"ntdll.dll") + GetProcAddress("wine_get_version")); only the code moved. I found each of the four patch sites in the new exe by matching the code around it in 5.0.0.1041720, with the RIP-relative displacements wildcarded. Each one matches exactly once, and the disassembly around it is unchanged:

Patch 1041720 1044392
slInit under Wine (DLSS SR/RR, Reflex) 0x1B709CB 0x1B7169B
DLSS-G "unsupported" on the Wine flag 0x1B759A0 0x1B76620
clear the cached Wine flag (RT, PT, FSR FG) 0x1EE2A53 0x1EDE8F3
buffer alignment as under Wine 0x1EE09F7 0x1EDC897
  • It supports both builds. It picks the build by exe size and still checks the sha256 (1044392: 9406eccc…fe51), so it refuses anything else.
  • Backups are now named per version (witcher3.exe.orig-5.0.0.1044392). Delete the old .orig-5.0.0.1041720 backup once you're on the new build.
  • NVIDIA: the DLSS patches were ported the same way but are untested on 1044392. Reports welcome.
  • GOG: I haven't checked whether the new GOG exe matches; the script will refuse it if it doesn't.

Close the game, then:

python3 w3_proton_patch.py --rt-only --dry-run   # should show "witcher3.exe 5.0.0.1044392" and all sites "original"
python3 w3_proton_patch.py --rt-only             # AMD/Intel
python3 w3_proton_patch.py --restore             # undo

Please don't use the workaround of commenting out the size/sha256 checks in the old script. On the new exe the old offsets land in the middle of unrelated instructions (a call target, a movss, an add rsp), so it corrupts code instead of patching the Wine check.

SSaancreed 2026-10-01 github

Friendly reminder that everyone can still fool the Wine detection check using simple

sed -i s/wine_get_version/frog_get_version/ witcher3.exe

executed in the same directory as the game's binary. No need for any fancy patchers that attempt to modify CPU instructions and can randomly break with updates. Though for anyone not on NVIDIA, it's recommended to use Proton Experimental bleeding-edge or set PROTON_DISABLE_NVAPI=1 for the title (see https://github.com/ValveSoftware/Proton/commit/e4bd4fc1fc5048a773674b7f617a9e1e1b8eb9a0).

Aarthuralvesdev 2026-10-01 github

Friendly reminder that everyone can still fool the Wine detection check using simple

sed -i s/wine_get_version/frog_get_version/ witcher3.exe
executed in the same directory as the game's binary. No need for any fancy patchers that attempt to modify CPU instructions and can randomly break with updates. Though for anyone not on NVIDIA, it's recommended to use Proton Experimental bleeding-edge or set PROTON_DISABLE_NVAPI=1 for the title (see e4bd4fc).

Thanks, worked!!!

IImanolFotia 2026-10-01 github

Replying to https://github.com/ValveSoftware/Proton/issues/607#issuecomment-5934737283

How did you make it work?
I tried it, but it would crash when loading any save
Tried lots of versions of Proton

Aarthuralvesdev 2026-10-01 github

Replying to https://github.com/ValveSoftware/Proton/issues/607#issuecomment-5935997210

Hi @ImanolFotia,

In my case, it worked without any crashes when loading saves. Here is exactly what I did and my setup:

  1. GPU: NVIDIA (RTX 4070).
  2. Proton Version: Proton Experimental.
  3. Patch applied: I ran the command specifically on the DX12 executable inside The Witcher 3/bin/x64_dx12/:
    sed -i s/wine_get_version/frog_get_version/ witcher3.exe
NNotYourAlejandro 2026-10-01 github

Lots of surfaces aren't rendering with patch v5.0.0c, not sure if this happened in the previous version.

https://github.com/user-attachments/assets/044bfa46-e633-4f3b-aa33-86f73947e3ab

Unfortunately the sed patch got me stuck on a black screen at launch, so no luck there.

EDIT: System info and diagnostics are here

EDIT 2: I had the HD Reworked mod installed, so my rendering issue is invalid.

BBlisto91 2026-10-01 github

@NotYourAlejandro Hi there. Could i get you to post a Proton log?
Also if you haven't already then i would try with Proton Experimental instead of 11.0-2

Ggabrielmaialva33 2026-10-01 github

FWIW on the sed one-liner failures (@NotYourAlejandro's black screen, @ImanolFotia's crash on load): hiding wine_get_version makes the game take its non-Wine path for everything, including requesting 256-byte-aligned buffers with D3D12_RESOURCE_FLAG_USE_TIGHT_ALIGNMENT. vkd3d-proton only accepts those since HansKristian-Work/vkd3d-proton#3308 (Sep 21), so on Proton 11.0-2 or GE-Proton11-7 the game probably dies at startup or on the first load. Proton Experimental should be fine for that part. With RT or path tracing on, NVIDIA also needs #3332 (in bleeding-edge) and AMD the dxil-spirv fix.

For anyone who'd rather stay on an older Proton: the patch script now supports 5.00c (5.0.0.1044392). Its default mode only unlocks DLSS/DLSS-G and keeps the Wine-side buffer alignment, so it works there: https://gist.github.com/gabrielmaialva33/33ebb2542f0513d55100b22aa2149ff5

NNotYourAlejandro 2026-10-01 github

@NotYourAlejandro Hi there. Could i get you to post a Proton log? Also if you haven't already then i would try with Proton Experimental instead of 11.0-2

Sure, here's the log using Proton Experimental.

EDIT: This log also captures the loooooong loading time when loading my save (ported from v4)

Aalasky17 maintainer 2026-10-01 github

@ImanolFotia Have you tried the bleeding-edge beta branch for experimental? That is needed for now, although should be updated on regular experimental by tomorrow afternoon :) For now, you need to look at the properties for the Proton - Experimental tool in Steam and select the bleeding-edge beta branch.

NNotYourAlejandro 2026-10-01 github

Just tried Proton Experimental bleeding edge, but didn't seem to change anything for me. here's my log: steam-292030.log

EDIT: I had the HD Reworked mod installed, so my rendering issue is invalid.

Ccarlosmsalado 2026-10-01 github

NVIDIA HairWorks is inert under Proton — it never loads its DLLs

(Separate issue from the DLSS/RT one above. Posting here because I don't see it reported for the Remastered — tell me if it belongs elsewhere.)

Setup

  • The Witcher 3: Wild Hunt — Remastered, Steam AppID 292030, exe 5.0.0.1044392 (build 25646871)
  • RTX 3080 (GA102), NVIDIA 610.57.04, kernel 7.2.5, Hyprland/Wayland, GE-Proton11-7
  • All four community exe patches applied (slInit + DLSS-G + Wine flag + Wine buffer alignment), i.e. the anti-Wine gate is cleared: RT/PT options do become selectable (verified) and DLSS works. So this is not the “NVIDIA features greyed out” bug — those are unlocked here.

Symptom
Selecting NVIDIA HairWorks → On (3rd entry, HairWorksLevel=2) is accepted, but the game silently drops it back to 0, and none of the three HairWorks DLLs is ever loaded:

  • NVHair_x64.dll, NVHairExt_x64.dll, GFSDK_HairWorks.win64.dll — all present in bin/x64_dx12/, none of them in the +loaddll stream of a PROTON_LOG=1 capture
  • the string hair does not appear once in 90,000+ log lines
  • no Xid, no VK_ERROR_DEVICE_LOST — nothing crashes; the feature simply never runs
  • visually: the game's default hair, not GameWorks

Ruled out, with evidence

  1. Not the anti-Wine flag. With the Wine-flag patches applied the HairWorks setting is still dropped, while RT/PT do become selectable. So it is a different gate.

  2. Not missing DLL dependencies. Import tables parsed:

    • NVHair_x64.dll → d3dx11_43.dll, KERNEL32.dll
    • NVHairExt_x64.dll → KERNEL32.dll, ADVAPI32.dll
    • GFSDK_HairWorks.win64.dll → D3DCOMPILER_47.dll, KERNEL32.dll

    Wine's builtin d3dx11_43.dll does export D3DX11CreateTextureFromMemory, the only symbol NVHair_x64.dll imports from it, and the game ships its own d3dcompiler_47.dll.

  3. Not the settings file, and not Steam Cloud. dx12user.settings is written on exit with HairWorksLevel=2 intact; for this app remotecache.vdf only lists The Witcher 3/gamesaves/* — the settings file is not cloud-synced at all.

  4. Not a UI capability gate. The HairWorks var in bin/config/r4game/user_config_matrix/pc/graphics.xml carries no enableIf…(Supported) tag — only refreshEngine;customNames;customPresetIfChanged;disableIfIntelGPU. The UI lets you pick it; the engine drops it afterwards.

Prime suspect: Linear Swept Spheres

LSS appears unsupported at every layer:

  • the game queries NvAPI_D3D12_GetRaytracingCaps (6/LINEAR_SWEPT_SPHERES) → Unsupported. dxvk-nvapi's GetRaytracingCaps implements only THREAD_REORDERING and OPACITY_MICROMAP.
  • vkd3d-proton: zero references to linear swept spheres, and no VK_NV_ray_tracing* extensions in libs/vkd3d/device.c.
  • the game's own graphics.xml gates the path-traced hair option on enableIfLinearSweptSpheresSupported + enableIfRTHairworksSupported + enableIfPTHair, which fits.
  • NVHairExt_x64.dll — 9.5 MB, the “extended” variant — points the same way.

Explicitly: this is inference, not proof. I can't see inside the engine. But it would explain why the option is offered, accepted, persisted briefly, and then silently discarded.

Related ground that already exists

  • vkd3d-proton CHANGELOG.md (next-gen update): "Hairworks is known to crash GPU on NVIDIA. More investigation is needed to root cause."

  • vkd3d-proton #1933 (Metro Exodus EE hairworks crash) is still open; the conclusion there was "it crashes inside the Hairworks DLL".

  • vkd3d-proton already fixes exactly this bug class for another title — libs/vkd3d/device_workarounds.c:

    /* Hair strand shaders write to a UAV, then read it back in the same workgroup, but misses a
     * device memory barrier in places, leading to GPU hang. */
    static const struct vkd3d_shader_quirk_info veilguard_quirks = {
        NULL, 0, VKD3D_SHADER_QUIRK_FORCE_DEVICE_MEMORY_BARRIER_THREAD_GROUP_COHERENCY,
    };
    

    witcher3_quirks has a single entry (the DXR VBO/CS barrier) and nothing for hair.

Ask

  1. Can anyone confirm whether the Remastered's HairWorks requires LSS (or another currently-stubbed NVAPI/DXR capability)? If it does, it might be worth documenting as “expected unavailable under Wine” rather than leaving people to guess.
  2. If it's the hair-strand-barrier class instead, could the Veilguard quirk reasonably be extended to TW3's hair shaders?

Happy to provide the full PROTON_LOG, dx12user.settings and the import-table dumps on request.

VValmar33 2026-10-02 github

Friendly reminder that everyone can still fool the Wine detection check using simple

sed -i s/wine_get_version/frog_get_version/ witcher3.exe

executed in the same directory as the game's binary. No need for any fancy patchers that attempt to modify CPU instructions and can randomly break with updates. Though for anyone not on NVIDIA, it's recommended to use Proton Experimental bleeding-edge or set PROTON_DISABLE_NVAPI=1 for the title (see e4bd4fc).

Sometimes, the sanest solutions are the simplest ones. :)

Kkisak-valve maintainer 2026-10-02 github

Witcher 3 Wild Hunt Remastered

Issue transferred from https://github.com/ValveSoftware/Proton/issues/10212.
@Renioo posted on 2026-10-02T09:28:55:

Compatibility Report

  • Name of the game with compatibility issues: Witcher 3 Wild Hunt Remastered
  • Steam AppID of the game: 292030

System Information

I confirm:

  • [ X] that I haven't found an existing compatibility report for this game.
  • [ X] that I have checked whether there are updates for my system available.

PROTON_ENABLE_WAYLAND=1 PROTON_ENABLE_HDR=1 VKD3D_FILTER_DEVICE_NAME="RTX 4070" DXVK_FILTER_DEVICE_NAME="RTX 4070" PROTON_ENABLE_NVAPI=1 %command%

Symptoms

When starting the game, the game will launch and play sound, however result in black screen. The issue is the game is defaulting output to integrated GPU, despite having NVIDIA GPU properly setup as first primary GPU. I did only encounter this issue with this title. Numerous same issues reported on REDDIT and solved by pointing to GPU as entered above.

Ggabrielmaialva33 2026-10-02 github

@NotYourAlejandro I had a look at both logs. They took different paths. In the Experimental one Streamline never starts (SL not initialized or no plugins found) and no RT state objects get built, which is the normal Wine path. In the bleeding-edge one Streamline, DLSS, DLSS-G, Reflex and DXR all load, which the game only does once Wine is hidden, and bleeding-edge doesn't hide it. Was the exe patched or sed'ed for that second run? Either way, if the missing walls/roofs and hair show up on both, the Wine check isn't the cause.

In your video whole meshes are gone (plaster walls, roofs, the hair mesh) while what's behind them draws fine, so they aren't being drawn at all. Neither log has a vkd3d error that explains that. I'd rule out mods first: does it also happen on a new game, not just the save from 4.x? Move mods/ and any non-CDPR folders in dlc/ aside, turn off mod.io mods in the menu, then Verify (it doesn't delete extra files in mods/).

If it's still there, try VKD3D_DISABLE_EXTENSIONS=VK_EXT_device_generated_commands %command%, then VKD3D_DISABLE_EXTENSIONS=VK_EXT_descriptor_buffer %command%, one at a time.


@carlosmsalado on the HairWorks part: I went through the 5.00c exe, and I don't think the DLL loads tell you anything. The DX12 exe has HairWorks built in (it even exports NvHairWorks_Create), and NVHair_x64, NVHairExt_x64 and GFSDK_HairWorks.win64 aren't referenced anywhere in it. They look like leftovers from the old DX11 build, so they don't load on Windows either. Its error callback is null as well, so no hair in the log is expected.

What gates it in the exe:

  • Regular HairWorks: only an Intel check on the adapter vendor (the engine side of disableIfIntelGPU). No Wine check reaches it.
  • Path-traced hair: the LSS cap from NvAPI_D3D12_GetRaytracingCaps, which dxvk-nvapi reports as none. So your LSS guess is right, for PT hair only.
  • Every preset in graphics.xml, RT and RT Ultra included, sets Virtual_HairWorksLevel=0, on Windows too. So picking a preset after turning HairWorks on switches it off again. Could that be what happened?

Here (RTX 4090, Proton Wineland, unmodified exe) it stays on across restarts and the hair renders.

NNotYourAlejandro 2026-10-02 github

I'd rule out mods first: does it also happen on a new game, not just the save from 4.x? Move mods/ and any non-CDPR folders in dlc/ aside, turn off mod.io mods in the menu, then Verify (it doesn't delete extra files in mods/).

Well this is awkward. It seems I forgot I had the HD Reworked mod installed which was the cause of the mesh and hair rendering issues; I usually play vanilla. Removing it fixed all demonstrated rendering issues. I appreciate the time you put into looking into the issue @gabrielmaialva33 and I apologize for wasting your time.

Aarthuralvesdev 2026-10-02 · hidden on GitHub github

I'd rule out mods first: does it also happen on a new game, not just the save from 4.x? Move mods/ and any non-CDPR folders in dlc/ aside, turn off mod.io mods in the menu, then Verify (it doesn't delete extra files in mods/).

Well this is awkward. It seems I forgot I had the HD Reworked mod installed which was the cause of the mesh and hair rendering issues; I usually play vanilla. Removing it fixed all demonstrated rendering issues. I appreciate the time you put into looking into the issue @gabrielmaialva33 and I apologize for wasting your time.

lol

Kkisak-valve maintainer 2026-10-02 github

The Witcher 3: Wild Hunt - Remastered (292030) - mostly in menus/cutscenes

Issue transferred from https://github.com/ValveSoftware/Proton/issues/10213.
@Snuffles20 posted on 2026-10-02T21:44:23:

Name of the game with compatibility issues: The Witcher 3: Wild Hunt - Remastered
Steam AppID of the game: 292030

System Information

  • GPU: AMD Radeon RX 9070 XT (Ryzen 7 9800X3D)
  • Video driver version: Mesa radv 26.2.4
  • Kernel version: 7.2.8-1-cachyos
  • Link to full system information report as Gist: https://gist.github.com/Snuffles20/223bc659bc55c46a5b9d5c0f230120ed
  • Proton version: Experimental 11.0-20260924, I also tried: (Hotfix, CachyOS Proton 11.0-20260703, GE-Proton11-7-x86_64])

I confirm:

  • [x] that I haven't found an existing compatibility report for this game.
  • [x] that I have checked whether there are updates for my system available.

Symptoms

The game keeps crashing on me, mostly in menus and cutscenes, and only rarely out in the world. The game's own "The Witcher 3 has crashed" window shows up. It doesn't happen at the same moment every time. Sometimes I can play for a long time in the menus and all is fine but suddenly it just crashes and I have to relaunch the game. I have become paranoid and keep saving every 2 minutes.

This isn't the DLSS/Streamline or black screen stuff in #10194 and #10212. I'm on AMD, and the game runs fine until it dies.

I got tired of guessing, so I tried to figure it out with some AI. The AI read the Proton logs. It seems to be the same crash every time: a null pointer read in witcher3.exe at +0x3317E3 (rcx=0). I've seen the same address three times, and the two logs I kept have an identical call chain (+0x1D4D0D6, +0x27E0D00, +0x1AC5903). It's a worker thread, and there's nothing from vkd3d, DXVK or the driver in the backtrace.

Things I've checked: the kernel log has no GPU errors, files verified on 5.00c, mods off, GPU at stock power, and I get the same result on other Proton builds. This is on KDE Plasma/Wayland.

Steam lists the game as Deck Verified, and my short test on a Steam Deck seemed fine, so it's something about desktop setups, or at least mine.

Sorry i am a bit of a noob when it comes to this technical stuff I have used AI to help me with this bug report. I hope everything is correct and helpful I am just trying to play the game. Let me know if you need more information or if i should refrain from posting more bugreports. :)

Reproduction

Just play. Menus and cutscenes are where it usually happens, and it can take a while to hit.
It did seem to happen multiple times for me in the cutscene when arriving to Royal Palace in Vizima. I had to skip the cutscene to get further into the game.

witcher3-5.00c-dx12-crash.log.gz

Log attached with PROTON_LOG=1 (my username is replaced with USER).

SSnuffles20 2026-10-02 github

I'm sorry is this really a duplicate. I can't find anyone else with the same issue. Were certain this should be closed as a duplicate?

Ccpuccino 2026-10-03 github

I'm sorry is this really a duplicate. I can't find anyone else with the same issue. Were certain this should be closed as a duplicate?

@Snuffles20 As mentioned in the ticket that was closed as duplicate "one issue report per title".

VValmar33 2026-10-03 github

@Snuffles20 Have you tried with the bleeding-edge branch of Proton Experimental? It includes an updated vkd3d-proton which might solve these issues.

SSnuffles20 2026-10-03 github

@Valmar33 Thanks for the suggestion! I hadn't thought about that. I'll test it out with the bleeding-edge branch and update the issue once I see if it fixes the problem or if it persists.

SSnuffles20 2026-10-03 github

@Valmar33 Tried the bleeding-edge branch as suggested. It still crashes, same bug.

Proton: experimental-bleeding-edge-11.0-445709-20261003
Kernel: 7.2.9-1-cachyos (updated since my first report, which was on 7.2.8-1)

Crash is identical to my earlier logs: EXCEPTION_ACCESS_VIOLATION, null read (rcx=0) at witcher3.exe + 0x3317E3, with the same callers (+0x1D4D0D6, +0x27E0D00, +0x1AC5903) and nothing from vkd3d, DXVK or the driver in the backtrace. That's the same call chain in all three logs I've saved, on two different Proton Experimental builds.

witcher3-proton-crash.log.26⁄10⁄04.gz

Eepistulae-ad-infernum-pro-te 2026-10-04 github

AMD RX 7900M NAVI 31 GFX1100

there is nothing.

https://gist.github.com/gabrielmaialva33/33ebb2542f0513d55100b22aa2149ff5

big problem with witcher 3 remastered.

The Witcher 3 Remastered (5.0) — Ray Tracing on AMD under Proton
AMD (RADV): dxil-spirv ( merged 2026-09-30 and in vkd3d-proton master). Without it, RADV crashes in vkCreateRayTracingPipelinesKHR while the game loads, because one of the game's path-tracing shaders translates to invalid SPIR-V ([Mesa #16450](https://gitlab.freedesktop.org/mesa/mesa/-/issues/16450))

ATTENTION: default value of option vk_wsi_force_swapchain_to_current_extent overridden by environment.
[Gamescope WSI] Application info:
  pApplicationName: (null)
  applicationVersion: 0
  pEngineName: (null)
  engineVersion: 0
  apiVersion: 4198400
[Gamescope WSI] Executable name: witcher3.exe
info:  Game: witcher3.exe
info:  DXVK: v3.1.1-55-gd30be2baea02a67
info:  Build: x86_64 gcc 14.0.0
info:  Vulkan: Found vkGetInstanceProcAddr in winevulkan.dll @ 0x6ffffbbf89a0
info:  Extension providers:
info:    Platform WSI
info:    OpenVR
info:    OpenXR
494101.168:03f4:03f8:trace:loaddll:build_module Loaded L"C:\\windows\\system32\\wineopenxr.dll" at 00006FFFEDDD0000: builtin
warn:  OpenXR: Unable to get required Vulkan instance extensions size
info:  Enabled instance extensions:
info:    VK_EXT_swapchain_colorspace
info:    VK_KHR_get_surface_capabilities2
info:    VK_KHR_surface
info:    VK_KHR_surface_maintenance1
info:    VK_KHR_win32_surface
ATTENTION: default value of option vk_wsi_force_swapchain_to_current_extent overridden by environment.
[Gamescope WSI] Application info:
  pApplicationName: witcher3.exe
  applicationVersion: 0
  pEngineName: DXVK
  engineVersion: 12587009
  apiVersion: 4206592

2 hours of trying different combinations:
Proton 11
Proton-experimental beta bleeding edge
Proton Hotfix
Proton cachyos
Proton GE 11
gamescope on/off
--skiplauncher
disable mods (obvious)

40GB download just to muck around with trying as many options and reading discussion forums & issues

and i still cant delete the 4GB soundtrack & artbook files out of the directory without them being replaced every update.

do i have a proton log? sure

Proton: 1791065757 experimental-bleeding-edge-11.0-445709-20261003-pe4bd4f-w750fd0-dd30be2-v31d1f8-x86_64
SteamGameId: 292030
Command: ['/home/amdr18/.local/share/Steam/steamapps/common/The Witcher 3/REDprelauncher.exe']
Options: {'disablenvapi', 'gamedrive', 'forcelgadd'}
depot: 4.0.20260805.254769
pressure-vessel: 0.20260805.0
pressure-vessel-arm64: 0.20260805.0
scripts: 0.20260805.0
steamrt4: 4.0.20260805.254769 steamrt4 4.0.20260805.254769
Kernel: Linux 7.2.7-amdr18 #1 SMP PREEMPT_DYNAMIC TKG Tue Sep 22 20:01:28 PDT 2026 x86_64
Language: LC_ALL None, LC_MESSAGES None, LC_CTYPE None
PATH: /home/amdr18/.local/share/Steam/steamapps/common/Proton - Experimental/files/bin/:/usr/bin:/bin
Effective WINEDEBUG: +timestamp,+pid,+tid,+seh,+unwind,+threadname,+debugstr,+loaddll,+mscoree

will it help? dont think so.

mesa-vulkan-drivers/resolute,now 26.2.3~kisak1~r amd64 [installed]
libvulkan1/now 1.4.357.0-1 amd64 [installed,local]

	apiVersion         = 1.4.354
	driverVersion      = 26.2.3
	vendorID           = 0x1002
	deviceID           = 0x744c
	deviceType         = PHYSICAL_DEVICE_TYPE_DISCRETE_GPU
	deviceName         = AMD Radeon RX 7900M (RADV NAVI31)
	driverID           = DRIVER_ID_MESA_RADV
	driverName         = radv
	driverInfo         = Mesa 26.2.3 - kisak-mesa PPA

mesa update required, wine proton fixes, and witcher3 fix to make anything work.
until then, i give up.

Ggabrielmaialva33 2026-10-04 github

@arrowgent the log you pasted stops at the header, so there's no way to tell what happened. Could you attach the whole steam-292030.log (PROTON_LOG=1 %command%) as a file, and say what "nothing" looks like: no window at all, a black screen, or a crash back to Steam? With bleeding-edge 20261003 both AMD fixes from the gist are already in, so whatever is left is something else.

Kkisak-valve maintainer 2026-10-04 github

The Witcher 3 Remaster The game freezes on the loading screen after dying.

Issue transferred from https://github.com/ValveSoftware/Proton/issues/10218.
@luisxx11 posted on 2026-10-04T13:19:54:

Compatibility Report

  • The game freezes on the loading screen after dying.
  • The Witcher 3
System Information
- RX 7600
- Mesa 26.1.99->
- Kernel version: SteamOS kernel 7.2
Fabricante da CPU: AuthenticAMD
Marca da CPU: AMD Ryzen 5 5500X3D 6-Core Processor

Driver: AMD AMD Radeon RX 7600 (radeonsi, navi33, ACO, DRM 3.65, 7.2.4-valve1-1-neptune-72-g5ab4af5e2eb9)
Versão do driver: 4.6 (Compatibility Profile) Mesa 26.1.2 (git-18b3910ac0)
Versão do OpenGL: 4.6
Profundidade de cor da área de trabalho: 24 bits por pixel
Taxa de atualização do monitor: 74 Hz
VendorID: 0x1002
DeviceID: 0x7480
Revisão não detectada
Quantidade de monitores: 1
Quantidade de placas de vídeo lógicas: 1
Resolução do monitor primário: 1920 x 1080
Resolução da área de trabalho: 1920 x 1080
Tamanho do monitor primário: 18.90" x 10.63" (21.65" diag), 48.0cm x 27.0cm (55.0cm diag)
VRAM primária: 8192 MB
Placa de som:
Dispositivo de áudio: ATI R6xx HDMI
Memória:
RAM: 32016 MB
  • Proton version: proton experimental

steam-292030.log

Ggabrielmaialva33 2026-10-04 github

@luisxx11 your log doesn't show a real hang: no crash, no GPU error, and the game keeps running until the end. The loading screen starts around 1058 s in the log and the game was closed about 40 s later.

There's a Steam thread about this exact thing on Steam Deck and desktop Linux: after dying, "load last save" takes 3–4 minutes, while loading the same save from the main menu takes a few seconds. One person there said switching to "Proton 11-02" brought it down to ~30 s: https://steamcommunity.com/app/292030/discussions/0/589565135028320986/

So for now, quit to the main menu and load from there. If you can, let it sit for ~5 minutes once and say whether it finishes. That would tell whether it's the same slow load or a real freeze.

Lluisxx11 2026-10-04 github

@gabrielmaialva33 Thanks bruh My bad, the loading takes a long time but it doesn't actually freeze. Thanks!

Lluisxx11 2026-10-04 github
Pperzeroo 2026-10-04 github

@luisxx11 I am not sure what window/desktop manager you're using and whether it will work the same, but on hyprland, I can switch to a workspace with steam on it, wait a few seconds and the loading screen gets past the hanging point and loads quicker. I am not sure why I need to switch to workspace with specifically steam, but it seems to be the fastest way, it boils down to the game just waiting for the render thread until it can load that specific part, but not sure why.

Lluisxx11 2026-10-04 github

@perzeroo I'm using Valve's SteamOS, I can't change its KDE environment, and even if I'm not in the gamescope UI, the loading is really slow.

Eepistulae-ad-infernum-pro-te 2026-10-04 github

@arrowgent the log you pasted stops at the header, so there's no way to tell what happened. Could you attach the whole steam-292030.log (PROTON_LOG=1 %command%) as a file, and say what "nothing" looks like: no window at all, a black screen, or a crash back to Steam? With bleeding-edge 20261003 both AMD fixes from the gist are already in, so whatever is left is something else.

i have even tried
proton-wineland-11.0-20260930-x86_64_v3

here is the log

steam-292030.log

"nothing happens"
after clicking Play on the launcher:
i can see witcher3.exe is running with 0% cpu 0% gpu and 0M of vram/ram
nothing appears. after 30 minutes. nothing.
after 5-10 minutes i just stop the game through steam

any other game with proton-experimental, proton-cachyos, proton 10
they all work fine.

witcher 3 4.0 next-gen worked fine previously

no modification to the witcher 3 files, pure vanilla out of the box.
mods are disabled in launcher, mods are unchecked in settings.
if i have to go somewhere to DELETE mods, idk...

at this point i wont be adding any more input or testing
reverting back to next-gen with cachyos & optiscaler.

thanks for the testing & help

Fflaax-io 2026-10-05 github

Replying to https://github.com/ValveSoftware/Proton/issues/607#issuecomment-5962070871

Same crash as @Snuffles20 (#10213, transferred above), with a reliable repro and a lot of isolation testing.

Summary

The Witcher 3 Remastered 5.00c (build 5.0.1044392) crashes with a null read in a render worker thread, always at the same address, on every Proton build and every setting I tried. Simplest repro: open the inventory and wait. I could narrow it down to what looks like a race inside the game (frame context reset vs. a pending scene culling job). More than 20 crashes, all identical.

Environment

  • CachyOS, kernel 7.2.9-1-cachyos, KDE Plasma (Wayland session), AMD Radeon RX 9070 XT (Mesa 26.2.4, RADV), Ryzen 7 9800X3D, 32 GB RAM, 2560x1440@144.
  • Game: Steam, build 5.0.1044392, witcher3.exe SHA-256 9406eccc12b68e08920931442ef6a57340e910d3e01f2082e88232487433fe51.

Crash signature (identical every time)

EXCEPTION_ACCESS_VIOLATION (0xC0000005, read at address 0x0) at witcher3.exe+0x3317e3 in thread RenderDispatcher2.
Call chain: witcher3+0x3317e3 <- +0x1d4d0d6 <- +0x27e0d00 <- +0x1ac5903 (same as in #10213).
The exception record is present in the minidump written by the game's crash reporter (Witcher3.dmp, exception stream: code 0xC0000005, address 0x1403317e3, parameters [0x0, 0x0]).

Reproduction

  1. Load a save (or start a new game) near Crossroads in White Orchard.
  2. Open the inventory (Tab) and leave it open without touching anything.
  3. The game crashes after roughly 1 to 12 minutes (shortest 39 s, longest 62 min). Switching inventory tabs or Alt-Tab makes it come sooner.
    It never happened in plain gameplay or in the ESC menu (world paused, picture frozen). Only menus with a live 3D scene show it: inventory, alchemy, glossary, character, potions, quest items. @Snuffles20 mentions menus and cutscenes, which fits.

What I tested (every row is a separate run, crash address unchanged in all)

Variable Variants tested Result
Proton build Valve Proton Experimental (Steam default, steamrt4), proton-cachyos 11.0-20261005 (official), wineland-11.0-20260930, GE-Proton11-1 all crash; Valve 340 s, proton-cachyos 39 s / 60 s / 78 s, GE 112 s, Wineland 91 s to 62 min
Install state pristine Steam install (Verify integrity of game files found nothing, no mods, default settings, new game), a heavily modded install (58 mods incl. a script merge), mods disabled in mods.settings, mod/DLC folders removed (verified via /proc/PID/fd that no mod file is opened) all crash
Executable unmodified exe; exe with a two-spot patch that only disables the game's wine_get_version check (needed for RT on Proton) both crash
Ray tracing on, off both crash (off: 540 s, 281 s)
Upscaler / AA FSR quality, AA off (no upscaler) both crash (AA off: 62 min)
Frame generation / latency FSR frame generation on/off, AMD Anti-Lag on/off, ENABLE_LAYER_MESA_ANTI_LAG=0 no effect
Sync ntsync, fsync, esync, plain wineserver (PROTON_NO_ESYNC=1 PROTON_NO_FSYNC=1 PROTON_NO_NTSYNC=1) all crash (plain wineserver: 1104 s)
Windowing X11 (Xwayland) vs native Wayland, borderless, exclusive fullscreen (237 s), windowed all crash
Frame pacing VSync on/off, game frame limit 144 and 60 (60 fps: 91 s; VSync off + 60: 330 s) no effect
CPU topology 8 logical CPUs, WINE_CPU_TOPOLOGY=4:0,1,2,3 no effect (162 s)
Input gamepad connected / not connected, rumble on/off no effect
GPU no GPU reset, no amdgpu kernel messages, GPU idle at the time of the freeze on older Wine builds not a driver problem
Wine crash handler older builds (wineland 20260930, GE-Proton11-1): game's exception filter blocks inside dbghelp!MiniDumpWriteDump and the window freezes; newer builds (proton-cachyos 20261005, Proton Experimental): the game's crash reporter dialog appears and the process exits the "hang" is just a blocked crash handler

What the code does

witcher3.exe+0x3317d0 is a short job body (49 bytes):

mov  rax,[rcx]            ; ctx
mov  rcx,[rax+0x4dd8]     ; job object
mov  rax,[rcx]            ; <-- faults, rcx == 0
call [rax+0x18]           ; virtual call on the job object
...
xchg [rcx+8],3            ; job state = done
  • The field ctx+0x4dd8 is assigned a freshly allocated 0x140-byte job object in a routine that also references the string SceneCullingTasks (and dispatches the job through the job system).
  • Other code zeroes the field while the next frame context is prepared (mov [rbx+0x4dd8],0 around +0x1bf30db). At the crash the whole context (about 20 KB) is zero-initialised.
  • So a culling job of the previous frame runs after its frame context was already reset. A quick experiment with a NULL check in the job body only moved the crash: the next one was at call [rax+0x18] (+0x3317eb) with a non-NULL but already destroyed object.
  • Everything points to a race inside the game (frame context reuse without waiting for the job). It is timing dependent, and my numbers suggest newer Wine builds hit it earlier (39 to 78 s with proton-cachyos 20261005 against minutes to an hour with Wineland/GE).

Observations that may help

  • Not specific to one Proton fork: same crash on Valve's own Proton Experimental.
  • Not caused by mods, RT, FSR, frame generation, sync primitives, windowing, frame limiter or the controller.
  • Freezes in the inventory, map and bestiary on the DX12 version were already reported in this thread in 2024 and 2025 (for example by zaps166, rariema and CrashTD, with RX 6600/6900 XT and also with frame limiters). It could be the same race, now more frequent on Remastered. I cannot prove that because those reports have no crash address.
  • Steam lists the game as Deck Verified. A short test on a Steam Deck is reported fine in #10213. Faster desktop hardware may widen the race window, but I can't tell.

Reports

  • proton-cachyos / Wineland: https://github.com/nanomatters/proton-cachyos/issues/91
  • I can also send the game's own crash report to CDPR (the dialog contains the exception record).
  • Disassembly of the function, backtraces of all threads (winedbg bt all) and the exception data are available as text. I'd rather share the minidumps privately.

Two days of testing later: I'd really like to go back to modding and playing. :,-) Happy to test anything else that helps.

SSnuffles20 2026-10-05 github

@flaax-io Damn bro you're insane and I feel validated, I'm not alone!!! No but seriously would be great to get a fix to this issue it runs so great otherwise. I had to download Windows on a new harddrive to play the game and I don't like it here.

Pp-valentin 2026-10-06 github

Fix for the random crash in 5.00c (inventory / menus): witcher3.exe+0x3317E3

If The Witcher 3 Remastered crashes on Linux or Steam Deck with "The Witcher 3 has crashed", mostly while the inventory, map, alchemy or character screen is open, it's probably this bug. Below is a small fix you can apply yourself. It's a bug in the game, not in Proton, and I found a way to patch it.

Results on my PC: before the fix the game crashed 4 times, the last two within 1–2 minutes. After the fix: no crashes so far, with the same FPS.

How to apply it (about 2 minutes)

  1. Close the game.
  2. Copy the code in "The fix script" below. Paste it into a text editor and save it as w3_culling_race_fix.py in your Downloads folder.
  3. Open a terminal. On Steam Deck, switch to Desktop Mode and open Konsole.
  4. Paste this line and press Enter:
    python3 ~/Downloads/w3_culling_race_fix.py
    
  5. It should say "Done: the crash fix is applied." Start the game from Steam as usual.

The script finds the game by itself. It works with normal or Flatpak Steam, the Steam Deck, and games installed on an SD card or a second drive.

  • Is it applied? Run the same line with --check at the end.
  • Undo: run the same line with --revert at the end. Or in Steam: right-click the game → Properties → Installed Files → Verify integrity of game files.
  • Your saves are never touched. The script keeps a backup of the original game file.
  • After a game update the fix is gone. That's expected: the update replaces the file. The script only works on version 5.00c. On any other version it changes nothing and tells you so.
The fix script (w3_culling_race_fix.py)
#!/usr/bin/env python3
"""
Crash fix for The Witcher 3: Wild Hunt - Remastered 5.00c on Linux / Steam Deck (Proton).

Fixes the random crash ("The Witcher 3 has crashed") that happens mostly while the
inventory, map, alchemy or character screen is open. Crash address: witcher3.exe+0x3317e3.

    python3 w3_culling_race_fix.py            apply the fix (finds the game automatically)
    python3 w3_culling_race_fix.py --check    only show whether the fix is applied
    python3 w3_culling_race_fix.py --revert   undo the fix
    python3 w3_culling_race_fix.py --revert "/path/to/bin/x64_dx12/witcher3.exe"   (manual path)

Close the game before running it. Saves are never touched. A backup of the original
game file is kept, and Steam's "Verify integrity of game files" also undoes the fix.
Only works on game version 5.00c (5.0.0.1044392); a game update removes the fix.

Technical summary (race in the game's culling-job cleanup):
  * Each frame context queues a scene-culling closure (+0x3317d0) on a worker thread.
    The closure reads ctx->job (ctx+0x4dd8), runs it and sets job->state (+0x8) = 3.
  * If the worker starts late, the main thread finishes the job inline and *forces*
    job->state = 3 (frame cleanup +0x1bf164c, "finish now" +0x1beafbe). The cleanup
    then frees the job and sets ctx->job = NULL; the stale closure runs and crashes.
  * Patch: "finish now" no longer forces state = 3, and the cleanup waits (PAUSE loop,
    ~60 ms cap, then original behaviour) until the closure has set state = 3 itself.
    The wait code lives in unused int3 padding at +0x21d8b08.
"""
import glob, hashlib, os, re, shutil, struct, sys

ORIG_SHA = "9406eccc12b68e08920931442ef6a57340e910d3e01f2082e88232487433fe51"
BASE = 0x140000000
SITE_A = 0x141bf164c      # cleanup: mov rcx,[rbx+4dd8]; mov eax,3; xchg [rcx+8],eax
SITE_A_RET = 0x141bf165b
SITE_B = 0x141beafbe      # finish-now: same 15 bytes
CAVE = 0x1421d8b08        # inside int3 padding 0x1421d8b03..0x1421d8b2f

FORCED = bytes.fromhex("488b8bd84d0000" "b803000000" "874108")  # 15 bytes
NOP15 = bytes.fromhex("0f1f840000000000" "0f1f8000000000")
GAME_EXE = os.path.join("steamapps", "common", "The Witcher 3", "bin", "x64_dx12", "witcher3.exe")


def rel32(src_next, dst):
    return struct.pack("<i", dst - src_next)


def cave_code():
    c = bytearray()
    c += bytes.fromhex("488b8bd84d0000")      # mov rcx,[rbx+0x4dd8]
    c += bytes.fromhex("b800004000")          # mov eax,0x400000   (spin budget, ~60 ms)
    c += bytes.fromhex("83790803")            # loop: cmp dword [rcx+8],3
    c += bytes.fromhex("7406")                #   je done
    c += bytes.fromhex("f390")                #   pause
    c += bytes.fromhex("ffc8")                #   dec eax
    c += bytes.fromhex("75f4")                #   jne loop
    c += bytes.fromhex("b803000000")          # done: mov eax,3
    c += bytes.fromhex("874108")              #   xchg [rcx+8],eax
    c += b"\xe9" + rel32(CAVE + len(c) + 5, SITE_A_RET)  # jmp back
    assert len(c) == 37
    return bytes(c)


def site_a_code():
    return b"\xe9" + rel32(SITE_A + 5, CAVE) + NOP15[:10]


def file_offset(data, va):
    pe = struct.unpack_from("<I", data, 0x3C)[0]
    nsec = struct.unpack_from("<H", data, pe + 6)[0]
    opt = struct.unpack_from("<H", data, pe + 20)[0]
    so = pe + 24 + opt
    rva = va - BASE
    for i in range(nsec):
        vsz, sva, rsz, raw = struct.unpack_from("<IIII", data, so + 40 * i + 8)
        if sva <= rva < sva + max(vsz, rsz):
            return raw + rva - sva
    raise ValueError(hex(va))


def find_game():
    """Look in the usual Steam locations (normal, Flatpak, Snap, Steam Deck) and
    in every extra Steam library listed in libraryfolders.vdf."""
    home = os.path.expanduser("~")
    roots = [
        f"{home}/.local/share/Steam",
        f"{home}/.steam/steam",
        f"{home}/.steam/root",
        f"{home}/.var/app/com.valvesoftware.Steam/.local/share/Steam",
        f"{home}/snap/steam/common/.local/share/Steam",
    ]
    libraries = []
    for root in roots:
        libraries.append(root)
        vdf = os.path.join(root, "steamapps", "libraryfolders.vdf")
        if os.path.isfile(vdf):
            libraries += re.findall(r'"path"\s+"([^"]+)"', open(vdf, errors="replace").read())
    libraries += glob.glob("/run/media/*/*") + glob.glob("/run/media/*")  # SD cards on Deck
    found = []
    for lib in libraries:
        exe = os.path.join(lib, GAME_EXE)
        if os.path.isfile(exe):
            real = os.path.realpath(exe)
            if real not in found:
                found.append(real)
    return found


def game_is_running():
    for pid in os.listdir("/proc"):
        if pid.isdigit():
            try:
                if b"witcher3.exe" in open(f"/proc/{pid}/cmdline", "rb").read():
                    return True
            except OSError:
                pass
    return False


def main():
    args = [a for a in sys.argv[1:] if not a.startswith("--")]
    revert, check = "--revert" in sys.argv, "--check" in sys.argv

    if args:
        exe = args[0]
    else:
        found = find_game()
        if not found:
            sys.exit("Could not find The Witcher 3. Run it again with the path to\n"
                     "bin/x64_dx12/witcher3.exe inside the game folder, in quotes.")
        if len(found) > 1:
            sys.exit("Found more than one installation; run it again with one of these paths:\n  "
                     + "\n  ".join(found))
        exe = found[0]
    print("Game file:", exe)

    backup = exe + ".orig-1044392"
    data = bytearray(open(exe, "rb").read())
    sha = hashlib.sha256(data).hexdigest()
    try:
        oa, ob, oc = (file_offset(data, x) for x in (SITE_A, SITE_B, CAVE))
        patched = (data[oa:oa + 15] == site_a_code() and data[ob:ob + 15] == NOP15
                   and data[oc:oc + 37] == cave_code())
    except (ValueError, struct.error):
        patched = False

    if check:
        if patched:
            print("Status: the crash fix IS applied.")
        elif sha == ORIG_SHA:
            print("Status: NOT applied (game version 5.00c, the fix can be applied).")
        else:
            print("Status: NOT applied, and this game version is not supported by this fix.")
        return

    if (revert or not patched) and game_is_running():
        sys.exit("The Witcher 3 is running. Close the game first, then run this again.")

    if revert:
        if not patched:
            print("The fix is not applied, nothing to undo.")
        elif os.path.exists(backup):
            shutil.copy2(backup, exe)
            print("Done: the original game file is restored.")
        else:
            sys.exit("No backup found. In Steam use: right-click the game > Properties >\n"
                     "Installed Files > Verify integrity of game files.")
        return

    if patched:
        print("The fix is already applied. Nothing to do.")
        return
    if sha != ORIG_SHA:
        sys.exit("This game version is not supported (only 5.00c, build 5.0.0.1044392).\n"
                 "If the game was updated, CD Projekt may have fixed the crash already.")
    if not (data[oa:oa + 15] == FORCED and data[ob:ob + 15] == FORCED
            and data[oc - 5:oc + 37 + 3] == b"\xcc" * 45):
        sys.exit("The game file does not look as expected; nothing was changed.")

    shutil.copy2(exe, backup)
    data[oa:oa + 15] = site_a_code()
    data[ob:ob + 15] = NOP15
    data[oc:oc + 37] = cave_code()
    tmp = exe + ".tmp"
    with open(tmp, "wb") as f:
        f.write(data)
    shutil.copymode(backup, tmp)
    os.replace(tmp, exe)
    print("Done: the crash fix is applied. Start the game from Steam as usual.")
    print("Backup of the original file:", backup)


if __name__ == "__main__":
    main()
Technical details (for the curious and for CDPR)

Builds on @flaax-io's analysis (#10213 / proton-cachyos#91). Exe 5.0.0.1044392, SHA-256 9406eccc…fe51. Tested on an RX 6700 XT and a Ryzen 5 5600X, with CachyOS, proton-cachyos and KDE Wayland.

Root cause

  • The per-frame dispatcher (+0x1bf1a40, plus a second one at +0x1c13a8c) allocates a culling job, stores it at ctx+0x4dd8, and queues a closure (+0x3317d0) on a worker.
  • The closure captures ctx, not the job. When it runs, it reads ctx->job, calls job->vfunc[0x18] (= sub_141bf2e70(job, 6), which runs the stages), and then sets job->state (+0x8) to 3.
  • If the worker hasn't started the closure yet, the main thread finishes the job inline with sub_141bf2e70(job, 6). It then forces job->state = 3 itself, in two places: the frame cleanup sub_141bf1610 (at +0x1bf164c) and "finish now" sub_141beaf90 (at +0x1beafbe).
  • The cleanup trusts state == 3. It releases the job and sets ctx+0x4dd8 = NULL (at +0x1bf177b). Both the destructor (+0x1bf3150) and the per-frame reset (+0x1c15149) go through this cleanup.
  • The closure that is still queued then runs and reads ctx->job == NULL, which is the crash at +0x3317e3. Or it reads a stale job, which explains the +0x3317eb crash after adding a NULL check. Wine's thread scheduling makes "the worker starts late" far more common than on Windows.

Patch (61 bytes)

  • At +0x1beafbe the forced state = 3 is NOPed, so only the closure marks the job done.
  • At +0x1bf164c the cleanup jumps to a stub in int3 padding at +0x21d8b08. The stub waits in a pause loop until the closure has set state == 3. The wait is capped at about 60 ms and then falls back to the original forced store, so it can't deadlock.
  • All other +0x4dd8 / state readers only do "if state != 3, help to stage N", which is idempotent.

A proper fix in the game would wait for the closure in the cleanup, or have the closure hold a reference to the job.

SSnuffles20 2026-10-06 github

@p-valentin Wow! I love smart people. I will try it now and let you know!

SSnuffles20 2026-10-06 github

@p-valentin I have to say it really seems to work. I haven't had extensive testing yet but I tried to brute force a crash like I manage to do very fast before and now It doesn't seem to happen. I will report back if it manages to crash as I continue to play.

Seriously, Thank you! for your work.

Pp-valentin 2026-10-07 github

@p-valentin I have to say it really seems to work. I haven't had extensive testing yet but I tried to brute force a crash like I manage to do very fast before and now It doesn't seem to happen. I will report back if it manages to crash as I continue to play.

Seriously, Thank you! for your work.

Please let me know. I also was unable to test it for longer than 1h.

Ggabrielmaialva33 2026-10-07 github

@p-valentin great find, and thanks for the write-up. A note for people using the DLSS/RT workaround from my gist: your script needs the untouched 5.00c exe, so it works as-is on Proton Wineland, which leaves the exe alone. With the sed one-liner, run your script first and the sed after. It can't be combined with my w3_proton_patch.py, which checks the original hash too. I'll point to your fix in the gist's troubleshooting table.

Ggabrielmaialva33 2026-10-07 github

@p-valentin one bug report for the script: game_is_running() matches "witcher3.exe" anywhere in any process's command line, so it also matches the script's own process (and the shell that started it) whenever the exe path is passed as an argument. That means the manual-path form, which GOG/Heroic, Faugus and Lutris installs need, always stops with "The Witcher 3 is running". Checking only the program each process runs fixes it; this is what I tested here (nothing running with a path argument → not running; a Proton-style S:\…\witcher3.exe process → running):

def game_is_running():
    for pid in os.listdir("/proc"):
        if not pid.isdigit() or int(pid) == os.getpid():
            continue
        try:
            argv = open(f"/proc/{pid}/cmdline", "rb").read().split(b"\0")[:2]
        except OSError:
            continue
        if any(re.split(rb"[\\/]", a)[-1].lower() == b"witcher3.exe" for a in argv):
            return True
    return False

Applied it on my 5.00c (Wineland, RTX 4090) through the auto-detected path and --check confirms it; no crash to report yet, but I haven't been able to trigger one on NVIDIA anyway.

Fflaax-io 2026-10-07 github

@p-valentin Confirmed here, thank you! Your fix works on my setup so far.

  • Before: crash at witcher3.exe+0x3317e3 after 39–78 s in the inventory with proton-cachyos 11.0-20261005.

  • After: 2+ hours without a crash, with inventory testing (tab switching, Alt-Tab, idling in menus), with RT, Frame Gen, Hairworks, 57 mods, and for half of that time Diablo 4 was running in parallel (just today). TW3 is still running.

I applied the patch to the untouched exe first and the two wine_get_version bytes for RT afterwards, as @gabrielmaialva33 suggested. The patch sites don't overlap.

Good job!

Ccrplcreek 2026-10-08 github

Confirmed here using Proton-CachyOS-Latest:

  • Before patch crashing in inventory menu with or without user input after a few minutes
  • No crashes since patch successfully applied. Tried to trigger crash via alt tabbing, moving through inventory menus, opening and closing inventory. Unable to reproduce crash after patching.

Operating System: CachyOS Linux
KDE Plasma Version: 6.7.5
KDE Frameworks Version: 6.30.0
Qt Version: 6.11.2
Kernel Version: 7.2.9-1-cachyos (64-bit)
Graphics Platform: Wayland
Processors: 6 × AMD Ryzen 5 7500X3D 6-Core Processor
Memory: 32 GiB of RAM (30.4 GiB usable)
Graphics Processor 1: AMD Radeon RX 9070
Graphics Processor 2: AMD Ryzen 5 7500X3D 6-Core Processor
Manufacturer: ASRock
Product Name: B850M-C

Ggabrielmaialva33 2026-10-08 github

@p-valentin patch 5.01 is out and your script refuses it (hash check), but the race is still there: the same 15-byte forced store at both sites. These are the 5.01 values (5.0.0.1048522, sha256 2df765398a3d1f165db418271be7200178320a4466817379f2fa0dc0001fb101) for your script:

ORIG_SHA = "2df765398a3d1f165db418271be7200178320a4466817379f2fa0dc0001fb101"
SITE_A = 0x141bf7f4c      # cleanup
SITE_A_RET = 0x141bf7f5b
SITE_B = 0x141bf189e      # finish-now
CAVE = 0x1421e1258        # inside int3 padding 0x1421e1253..0x1421e127f

With those (and a .orig-1048522 backup name) your script accepts the 5.01 exe, changes the same 61 bytes, --check confirms it and --revert brings the original sha back. I only checked that on a copy of the exe: I can't trigger the crash on NVIDIA, so it still needs a run from someone who could.

SSnuffles20 2026-10-08 github

@gabrielmaialva33 I'm sorry I am really trying to understand what you changed in the script to get it working again but I can't seem to figure it out. Would you be so kind as to break it down a little further. I can test if the changes works.

Thank you!

Ggabrielmaialva33 2026-10-08 github

@Snuffles20 sure, here it is ready to run: p-valentin's script with only the 5.01 values changed, so you don't have to edit anything.

Same steps as his: save it as w3_culling_race_fix_501.py in Downloads, then run python3 ~/Downloads/w3_culling_race_fix_501.py (add --check to see the status, --revert to undo). Run it on the untouched 5.01 exe; if you also use the sed one-liner or my script for RT, apply those after it.

What changed from his 5.00c version: the exe's sha256, where the two patch sites and the free space for the wait loop moved to in 5.01, the backup name, and game_is_running(), which no longer trips over its own command line, so a manual path works too. The 61 patch bytes are the same. I tested apply, --check and --revert on a copy of the 5.01 exe, but I can't trigger the crash on NVIDIA, so you're the real test. If @p-valentin posts his own 5.01 update, use his.

w3_culling_race_fix_501.py
#!/usr/bin/env python3
"""
Crash fix for The Witcher 3: Wild Hunt - Remastered 5.01 on Linux / Steam Deck (Proton).

p-valentin's fix (https://github.com/ValveSoftware/Proton/issues/607#issuecomment-6024287034), ported to
patch 5.01 (5.0.0.1048522): new sha256, SITE_A, SITE_A_RET, SITE_B, CAVE and backup name, and a
game_is_running() that no longer matches its own command line. The patch bytes are unchanged.

Fixes the random crash ("The Witcher 3 has crashed") that happens mostly while the
inventory, map, alchemy or character screen is open (witcher3.exe+0x3317e3 on 5.00c).

    python3 w3_culling_race_fix.py            apply the fix (finds the game automatically)
    python3 w3_culling_race_fix.py --check    only show whether the fix is applied
    python3 w3_culling_race_fix.py --revert   undo the fix
    python3 w3_culling_race_fix.py --revert "/path/to/bin/x64_dx12/witcher3.exe"   (manual path)

Close the game before running it. Saves are never touched. A backup of the original
game file is kept, and Steam's "Verify integrity of game files" also undoes the fix.
Only works on game version 5.01 (5.0.0.1048522); a game update removes the fix.

Technical summary (race in the game's culling-job cleanup; addresses below are 5.00c's):
  * Each frame context queues a scene-culling closure (+0x3317d0) on a worker thread.
    The closure reads ctx->job (ctx+0x4dd8), runs it and sets job->state (+0x8) = 3.
  * If the worker starts late, the main thread finishes the job inline and *forces*
    job->state = 3 (frame cleanup +0x1bf164c, "finish now" +0x1beafbe). The cleanup
    then frees the job and sets ctx->job = NULL; the stale closure runs and crashes.
  * Patch: "finish now" no longer forces state = 3, and the cleanup waits (PAUSE loop,
    ~60 ms cap, then original behaviour) until the closure has set state = 3 itself.
    The wait code lives in unused int3 padding at +0x21d8b08.
"""
import glob, hashlib, os, re, shutil, struct, sys

ORIG_SHA = "2df765398a3d1f165db418271be7200178320a4466817379f2fa0dc0001fb101"
BASE = 0x140000000
SITE_A = 0x141bf7f4c      # cleanup: mov rcx,[rbx+4dd8]; mov eax,3; xchg [rcx+8],eax (5.01)
SITE_A_RET = 0x141bf7f5b
SITE_B = 0x141bf189e      # finish-now: same 15 bytes
CAVE = 0x1421e1258        # inside int3 padding 0x1421e1253..0x1421e127f

FORCED = bytes.fromhex("488b8bd84d0000" "b803000000" "874108")  # 15 bytes
NOP15 = bytes.fromhex("0f1f840000000000" "0f1f8000000000")
GAME_EXE = os.path.join("steamapps", "common", "The Witcher 3", "bin", "x64_dx12", "witcher3.exe")


def rel32(src_next, dst):
    return struct.pack("<i", dst - src_next)


def cave_code():
    c = bytearray()
    c += bytes.fromhex("488b8bd84d0000")      # mov rcx,[rbx+0x4dd8]
    c += bytes.fromhex("b800004000")          # mov eax,0x400000   (spin budget, ~60 ms)
    c += bytes.fromhex("83790803")            # loop: cmp dword [rcx+8],3
    c += bytes.fromhex("7406")                #   je done
    c += bytes.fromhex("f390")                #   pause
    c += bytes.fromhex("ffc8")                #   dec eax
    c += bytes.fromhex("75f4")                #   jne loop
    c += bytes.fromhex("b803000000")          # done: mov eax,3
    c += bytes.fromhex("874108")              #   xchg [rcx+8],eax
    c += b"\xe9" + rel32(CAVE + len(c) + 5, SITE_A_RET)  # jmp back
    assert len(c) == 37
    return bytes(c)


def site_a_code():
    return b"\xe9" + rel32(SITE_A + 5, CAVE) + NOP15[:10]


def file_offset(data, va):
    pe = struct.unpack_from("<I", data, 0x3C)[0]
    nsec = struct.unpack_from("<H", data, pe + 6)[0]
    opt = struct.unpack_from("<H", data, pe + 20)[0]
    so = pe + 24 + opt
    rva = va - BASE
    for i in range(nsec):
        vsz, sva, rsz, raw = struct.unpack_from("<IIII", data, so + 40 * i + 8)
        if sva <= rva < sva + max(vsz, rsz):
            return raw + rva - sva
    raise ValueError(hex(va))


def find_game():
    """Look in the usual Steam locations (normal, Flatpak, Snap, Steam Deck) and
    in every extra Steam library listed in libraryfolders.vdf."""
    home = os.path.expanduser("~")
    roots = [
        f"{home}/.local/share/Steam",
        f"{home}/.steam/steam",
        f"{home}/.steam/root",
        f"{home}/.var/app/com.valvesoftware.Steam/.local/share/Steam",
        f"{home}/snap/steam/common/.local/share/Steam",
    ]
    libraries = []
    for root in roots:
        libraries.append(root)
        vdf = os.path.join(root, "steamapps", "libraryfolders.vdf")
        if os.path.isfile(vdf):
            libraries += re.findall(r'"path"\s+"([^"]+)"', open(vdf, errors="replace").read())
    libraries += glob.glob("/run/media/*/*") + glob.glob("/run/media/*")  # SD cards on Deck
    found = []
    for lib in libraries:
        exe = os.path.join(lib, GAME_EXE)
        if os.path.isfile(exe):
            real = os.path.realpath(exe)
            if real not in found:
                found.append(real)
    return found


def game_is_running():
    for pid in os.listdir("/proc"):
        if not pid.isdigit() or int(pid) == os.getpid():
            continue
        try:
            argv = open(f"/proc/{pid}/cmdline", "rb").read().split(b"\0")[:2]
        except OSError:
            continue
        if any(re.split(rb"[\\/]", a)[-1].lower() == b"witcher3.exe" for a in argv):
            return True
    return False


def main():
    args = [a for a in sys.argv[1:] if not a.startswith("--")]
    revert, check = "--revert" in sys.argv, "--check" in sys.argv

    if args:
        exe = args[0]
    else:
        found = find_game()
        if not found:
            sys.exit("Could not find The Witcher 3. Run it again with the path to\n"
                     "bin/x64_dx12/witcher3.exe inside the game folder, in quotes.")
        if len(found) > 1:
            sys.exit("Found more than one installation; run it again with one of these paths:\n  "
                     + "\n  ".join(found))
        exe = found[0]
    print("Game file:", exe)

    backup = exe + ".orig-1048522"
    data = bytearray(open(exe, "rb").read())
    sha = hashlib.sha256(data).hexdigest()
    try:
        oa, ob, oc = (file_offset(data, x) for x in (SITE_A, SITE_B, CAVE))
        patched = (data[oa:oa + 15] == site_a_code() and data[ob:ob + 15] == NOP15
                   and data[oc:oc + 37] == cave_code())
    except (ValueError, struct.error):
        patched = False

    if check:
        if patched:
            print("Status: the crash fix IS applied.")
        elif sha == ORIG_SHA:
            print("Status: NOT applied (game version 5.01, the fix can be applied).")
        else:
            print("Status: NOT applied, and this game version is not supported by this fix.")
        return

    if (revert or not patched) and game_is_running():
        sys.exit("The Witcher 3 is running. Close the game first, then run this again.")

    if revert:
        if not patched:
            print("The fix is not applied, nothing to undo.")
        elif os.path.exists(backup):
            shutil.copy2(backup, exe)
            print("Done: the original game file is restored.")
        else:
            sys.exit("No backup found. In Steam use: right-click the game > Properties >\n"
                     "Installed Files > Verify integrity of game files.")
        return

    if patched:
        print("The fix is already applied. Nothing to do.")
        return
    if sha != ORIG_SHA:
        sys.exit("This game version is not supported (only 5.01, build 5.0.0.1048522).\n"
                 "If the game was updated, CD Projekt may have fixed the crash already.")
    if not (data[oa:oa + 15] == FORCED and data[ob:ob + 15] == FORCED
            and data[oc - 5:oc + 37 + 3] == b"\xcc" * 45):
        sys.exit("The game file does not look as expected; nothing was changed.")

    shutil.copy2(exe, backup)
    data[oa:oa + 15] = site_a_code()
    data[ob:ob + 15] = NOP15
    data[oc:oc + 37] = cave_code()
    tmp = exe + ".tmp"
    with open(tmp, "wb") as f:
        f.write(data)
    shutil.copymode(backup, tmp)
    os.replace(tmp, exe)
    print("Done: the crash fix is applied. Start the game from Steam as usual.")
    print("Backup of the original file:", backup)


if __name__ == "__main__":
    main()
SSnuffles20 2026-10-08 github

@gabrielmaialva33 I can confirm it works after very short testing but I tried it on the new update without the script and I managed to crash it within 1-2 minutes. With the updated script. Thank you! very much by the way it works great. I tried to force a crash again and after spamming the menus straight for a few minutes no crash in sight. I will update if it does as I keep playing. Thank you for your time explaining the script in better detail. and of course a big thanks to @p-valentin for the orginal script. You guys are awesome and I couldn't do it without you.

Ggabrielmaialva33 2026-10-08 github

@Snuffles20 Glad it held up, thanks for testing it properly. I've linked it from the gist.

Ccoreybruce 2026-10-09 github

Ever since the free remastered update has come out I have been unable to even play the game, I get the launcher but after pressing play nothing happens , same thing happen when I do --launcher-skip. I have tried proton experiemental, Proton 11, ProtonGE and proton CachyOS, I even reinstalled the game with no change

https://gist.github.com/coreybruce/a1df988e51085aa35d6897ec8c1a7ebe

CPU: AMD Ryzen 7 5700X
Ram: 32GB
GPU: AMD Radeon RX 9070 XT

Pp-valentin 2026-10-09 github

@gabrielmaialva33 thanks for the 5.01 port, your values were spot on. And thanks @Snuffles20 for testing it properly, that's great confirmation.

So that nobody has to port offsets after every patch, I rewrote the script to find the code itself. It should keep working across game updates as long as the bug is there.

What changed

  • No fixed addresses and no version check. It locates the culling closure by its instruction pattern, reads the job offset from it (still 0x4dd8), finds the two forced state = 3 stores, and identifies the cleanup as the one whose function also clears the job pointer. It then picks
    What changed
  • No fixed addresses and no version check. It locates the culling closure by its instruction pattern, reads the job offset from it (still 0x4dd8), finds the two forced state = 3 stores, and identifies the cleanup as the one whose function also clears the job pointer. It then picks unused int3 padding between functions for the wait code.
  • It refuses rather than guesses. It only patches if it finds exactly two sites (one cleanup, one finish-now), both calling the same function. Otherwise it changes nothing and says why.
  • Compatible with the earlier scripts. It produces the same 61 bytes in the same places. On a game patched with the 5.00c script or Gabriel's 5.01 script, --check reports it as applied, and --revert works with their .orig-* backups.
  • Gabriel's game_is_running() fix for manual paths is included.

Tested

  • 5.00c: output byte-identical to the original patch.
  • 5.01 (2df76539…fb101): finds cleanup 0x141bf7f4c, finish-now 0x141bf189e, cave 0x1421e1258, the same as Gabriel's values.
  • 5.01 unpatched: crashed after ~18 min, null read at witcher3.exe+0x3302f3. That's the closure's job load in 5.01, the same bug moved.
  • 5.01 patched: about 1 hour of play, no crash. RX 6700 XT / Ryzen 5 5600X, CachyOS, proton-cachyos.

How to use it: close the game, save the script below as w3_culling_race_fix.py, then run python3 w3_culling_race_fix.py. Use --check to see the status and --revert to undo. After a game update, just run it again. No new version needed.

w3_culling_race_fix.py (works on 5.00c, 5.01 and, hopefully, later versions)
#!/usr/bin/env python3
"""
Crash fix for The Witcher 3: Wild Hunt - Remastered (5.x) on Linux / Steam Deck (Proton).

Fixes the random crash ("The Witcher 3 has crashed") that happens mostly while the
inventory, map, alchemy or character screen is open. Crash address in 5.00c:
witcher3.exe+0x3317e3.

    python3 w3_culling_race_fix.py            apply the fix (finds the game automatically)
    python3 w3_culling_race_fix.py --check    only show whether the fix is applied
    python3 w3_culling_race_fix.py --revert   undo the fix
    python3 w3_culling_race_fix.py "/path/to/bin/x64_dx12/witcher3.exe"   (manual path)

Close the game before running it. Saves are never touched. A backup of the original
game file is kept, and Steam's "Verify integrity of game files" also undoes the fix.

The fix does not depend on one game version: it searches the game file for the code
involved. After a game update simply run it again. If the code has changed so much that
it can't be found with certainty, the script changes nothing and says so.

Technical summary (race in the game's culling-job cleanup):
  * Each frame context queues a scene-culling closure on a worker thread. The closure
    reads ctx->job (ctx+X, X = 0x4dd8 in 5.00c), runs it and sets job->state (+0x8) = 3.
  * If the worker starts late, the main thread finishes the job inline and *forces*
    job->state = 3 (in the frame cleanup and in a "finish now" function). The cleanup
    then frees the job and sets ctx->job = NULL; the stale closure runs and crashes.
  * Patch: "finish now" no longer forces state = 3, and the cleanup waits (PAUSE loop,
    ~60 ms cap, then original behaviour) until the closure has set state = 3 itself.
    The wait code is placed in unused int3 padding between functions.
"""
import bisect, glob, hashlib, os, re, shutil, struct, sys

BASE_DEFAULT = 0x140000000
GAME_EXE = os.path.join("steamapps", "common", "The Witcher 3", "bin", "x64_dx12", "witcher3.exe")
BACKUP_SUFFIX = ".w3fix-original"
NOP15 = bytes.fromhex("0f1f840000000000" "0f1f8000000000")
WAIT_ARG = b"\xba\x06\x00\x00\x00"          # mov edx,6 (stage passed to the inline finish)


class NotFound(Exception):
    pass


# ---------------------------------------------------------------- PE helpers
class PE:
    def __init__(self, data):
        self.data = data
        if data[:2] != b"MZ":
            raise NotFound("not a Windows executable")
        pe = struct.unpack_from("<I", data, 0x3C)[0]
        nsec = struct.unpack_from("<H", data, pe + 6)[0]
        opt = struct.unpack_from("<H", data, pe + 20)[0]
        self.base = struct.unpack_from("<Q", data, pe + 24 + 24)[0] or BASE_DEFAULT
        self.secs = {}
        so = pe + 24 + opt
        for i in range(nsec):
            name = data[so + 40 * i:so + 40 * i + 8].rstrip(b"\0").decode(errors="replace")
            vsz, va, rsz, raw = struct.unpack_from("<IIII", data, so + 40 * i + 8)
            self.secs[name] = (va, vsz, raw, rsz)
        if ".text" not in self.secs or ".pdata" not in self.secs:
            raise NotFound("unexpected file layout")
        va, vsz, raw, rsz = self.secs[".text"]
        self.text_raw, self.text_va, self.text_len = raw, va, min(vsz, rsz)
        va, vsz, raw, rsz = self.secs[".pdata"]
        funcs = []
        for o in range(raw, raw + min(vsz, rsz) - 11, 12):
            b, e, _ = struct.unpack_from("<III", data, o)
            if b:
                funcs.append((b, e))
        funcs.sort()
        self.funcs, self.starts = funcs, [f[0] for f in funcs]

    def text(self):
        return bytes(self.data[self.text_raw:self.text_raw + self.text_len])

    def off2rva(self, off):            # offset into text() -> RVA
        return off + self.text_va

    def rva2off(self, rva):
        for va, vsz, raw, rsz in self.secs.values():
            if va <= rva < va + max(vsz, rsz):
                return raw + rva - va
        raise NotFound("address outside the file")

    def func_of(self, rva):
        i = bisect.bisect_right(self.starts, rva) - 1
        if i >= 0 and self.funcs[i][0] <= rva < self.funcs[i][1]:
            return self.funcs[i]
        return None


def rel32(src_next, dst):
    return struct.pack("<i", dst - src_next)


def forced_store(x):                  # mov rcx,[rbx+X]; mov eax,3; xchg [rcx+8],eax
    return b"\x48\x8b\x8b" + struct.pack("<I", x) + b"\xb8\x03\x00\x00\x00\x87\x41\x08"


def cave_code(x, cave_rva, ret_rva):
    c = bytearray(b"\x48\x8b\x8b" + struct.pack("<I", x))  # mov rcx,[rbx+X]
    c += bytes.fromhex("b800004000")          # mov eax,0x400000   (spin budget, ~60 ms)
    c += bytes.fromhex("83790803")            # loop: cmp dword [rcx+8],3
    c += bytes.fromhex("7406")                #   je done
    c += bytes.fromhex("f390")                #   pause
    c += bytes.fromhex("ffc8")                #   dec eax
    c += bytes.fromhex("75f4")                #   jne loop
    c += bytes.fromhex("b803000000")          # done: mov eax,3
    c += bytes.fromhex("874108")              #   xchg [rcx+8],eax
    c += b"\xe9" + rel32(cave_rva + len(c) + 5, ret_rva)
    assert len(c) == 37
    return bytes(c)


CAVE_LEN = 37
CAVE_BODY_RE = (rb"\x48\x8b\x8b(....)\xb8\x00\x00\x40\x00\x83\x79\x08\x03\x74\x06\xf3\x90"
                rb"\xff\xc8\x75\xf4\xb8\x03\x00\x00\x00\x87\x41\x08\xe9(....)")


# ---------------------------------------------------------------- analysis
def find_job_offset(text):
    """The closure body: mov rax,[rcx]; mov rbx,rcx; mov rcx,[rax+X]; mov rax,[rcx];
    call [rax+0x18]; mov rax,[rbx]; mov rcx,[rax+X]; mov eax,3; xchg [rcx+8],eax"""
    pat = re.compile(rb"\x48\x8b\x01\x48\x8b\xd9\x48\x8b\x88(....)\x48\x8b\x01\xff\x50\x18"
                     rb"\x48\x8b\x03\x48\x8b\x88(....)\xb8\x03\x00\x00\x00\x87\x41\x08", re.S)
    xs = set()
    for m in pat.finditer(text):
        a, b = m.group(1), m.group(2)
        if a == b:
            xs.add(struct.unpack("<I", a)[0])
    if len(xs) != 1:
        raise NotFound("could not identify the culling-job code (found %d candidates)" % len(xs))
    return xs.pop()


def analyse(pe):
    """Returns a dict describing the patch sites, for an unpatched or patched file."""
    text = pe.text()
    x = find_job_offset(text)
    xb = struct.pack("<I", x)

    # unpatched sites: call <finish>; mov rcx,[rbx+X]; mov eax,3; xchg [rcx+8],eax
    sites = []
    for m in re.finditer(re.escape(WAIT_ARG) + rb"\xe8(....)" + re.escape(forced_store(x)), text, re.S):
        off = m.start() + 10
        tgt = pe.off2rva(off) + struct.unpack("<i", m.group(1))[0]
        sites.append((pe.off2rva(off), tgt))
    # already patched sites
    patched_a, patched_b = [], []
    for m in re.finditer(re.escape(WAIT_ARG) + rb"\xe8(....)\xe9(....)" + re.escape(NOP15[:10]), text, re.S):
        rva = pe.off2rva(m.start() + 10)
        dst = rva + 5 + struct.unpack("<i", m.group(2))[0]
        o = pe.rva2off(dst)
        cm = re.fullmatch(CAVE_BODY_RE, bytes(pe.data[o:o + CAVE_LEN]), re.S)
        if cm and cm.group(1) == xb:
            patched_a.append((rva, dst))
    for m in re.finditer(re.escape(WAIT_ARG) + rb"\xe8(....)" + re.escape(NOP15), text, re.S):
        patched_b.append(pe.off2rva(m.start() + 10))

    def stores_x(rva):          # does the enclosing function store to [reg+X]? (= cleanup)
        f = pe.func_of(rva)
        if not f:
            return False
        body = bytes(pe.data[pe.rva2off(f[0]):pe.rva2off(f[0]) + (f[1] - f[0])])
        return re.search(rb"[\x48\x4c]\x89[\x80-\xbf]" + re.escape(xb), body, re.S) is not None

    return dict(x=x, sites=sites, cleanup=[s for s in sites if stores_x(s[0])],
                finish=[s for s in sites if not stores_x(s[0])],
                patched_a=patched_a, patched_b=patched_b)


def find_cave(pe):
    text = pe.text()
    for m in re.finditer(rb"\xcc{%d,}" % (CAVE_LEN + 8), text):
        start = (m.start() + 4 + 7) & ~7                # keep int3 before, align to 8
        if start + CAVE_LEN + 2 > m.end():
            continue
        rva = pe.off2rva(start)
        if pe.func_of(rva) or pe.func_of(rva + CAVE_LEN - 1):
            continue                                    # only padding between functions
        return rva
    raise NotFound("no free space found for the fix")


def status(pe):
    a = analyse(pe)
    if len(a["patched_a"]) == 1 and len(a["patched_b"]) == 1 and not a["sites"]:
        return "patched", a
    if (len(a["sites"]) == 2 and len(a["cleanup"]) == 1 and len(a["finish"]) == 1
            and a["sites"][0][1] == a["sites"][1][1] and not a["patched_a"]):
        return "unpatched", a
    raise NotFound("the game code does not match what this fix expects "
                   "(sites=%d, cleanup=%d, finish=%d, patched=%d/%d)"
                   % (len(a["sites"]), len(a["cleanup"]), len(a["finish"]),
                      len(a["patched_a"]), len(a["patched_b"])))


def build_patch(data):
    pe = PE(data)
    st, a = status(pe)
    if st != "unpatched":
        return None, a
    out = bytearray(data)
    x = a["x"]
    site_a = a["cleanup"][0][0]
    site_b = a["finish"][0][0]
    cave = find_cave(pe)
    oa, ob, oc = pe.rva2off(site_a), pe.rva2off(site_b), pe.rva2off(cave)
    assert out[oa:oa + 15] == forced_store(x) and out[ob:ob + 15] == forced_store(x)
    out[oa:oa + 15] = b"\xe9" + rel32(site_a + 5, cave) + NOP15[:10]
    out[ob:ob + 15] = NOP15
    out[oc:oc + CAVE_LEN] = cave_code(x, cave, site_a + 15)
    a.update(site_a=site_a, site_b=site_b, cave=cave)
    return out, a


# ---------------------------------------------------------------- environment
def find_game():
    """Usual Steam locations (normal, Flatpak, Snap, Steam Deck) and every extra
    Steam library listed in libraryfolders.vdf."""
    home = os.path.expanduser("~")
    roots = [
        f"{home}/.local/share/Steam",
        f"{home}/.steam/steam",
        f"{home}/.steam/root",
        f"{home}/.var/app/com.valvesoftware.Steam/.local/share/Steam",
        f"{home}/snap/steam/common/.local/share/Steam",
    ]
    libraries = []
    for root in roots:
        libraries.append(root)
        vdf = os.path.join(root, "steamapps", "libraryfolders.vdf")
        if os.path.isfile(vdf):
            libraries += re.findall(r'"path"\s+"([^"]+)"', open(vdf, errors="replace").read())
    libraries += glob.glob("/run/media/*/*") + glob.glob("/run/media/*")  # SD cards on Deck
    found = []
    for lib in libraries:
        exe = os.path.join(lib, GAME_EXE)
        if os.path.isfile(exe):
            real = os.path.realpath(exe)
            if real not in found:
                found.append(real)
    return found


def game_is_running():
    for pid in os.listdir("/proc"):
        if not pid.isdigit() or int(pid) == os.getpid():
            continue
        try:
            argv = open(f"/proc/{pid}/cmdline", "rb").read().split(b"\0")[:2]
        except OSError:
            continue
        if any(re.split(rb"[\\/]", a)[-1].lower() == b"witcher3.exe" for a in argv):
            return True
    return False


def find_backup(exe, current):
    """A backup whose patched form is exactly the current file."""
    cands = [exe + BACKUP_SUFFIX] + sorted(glob.glob(glob.escape(exe) + ".orig-*"))
    for b in cands:
        if os.path.isfile(b):
            try:
                patched, _ = build_patch(bytearray(open(b, "rb").read()))
            except NotFound:
                continue
            if patched is not None and patched == current:
                return b
    return None


def main():
    args = [a for a in sys.argv[1:] if not a.startswith("--")]
    revert, check = "--revert" in sys.argv, "--check" in sys.argv

    if args:
        exe = args[0]
    else:
        found = find_game()
        if not found:
            sys.exit("Could not find The Witcher 3. Run it again with the path to\n"
                     "bin/x64_dx12/witcher3.exe inside the game folder, in quotes.")
        if len(found) > 1:
            sys.exit("Found more than one installation; run it again with one of these paths:\n  "
                     + "\n  ".join(found))
        exe = found[0]
    print("Game file:", exe)

    data = bytearray(open(exe, "rb").read())
    try:
        st, info = status(PE(data))
    except NotFound as e:
        if check:
            print("Status: NOT applied. This game version can't be patched:", e)
            return
        sys.exit("Nothing was changed: %s.\nThe game may have been updated in a way this "
                 "fix doesn't support (or CD Projekt fixed the crash)." % e)

    if check:
        print("Status: the crash fix IS applied." if st == "patched" else
              "Status: NOT applied (this game version can be patched).")
        return

    if (revert or st == "unpatched") and game_is_running():
        sys.exit("The Witcher 3 is running. Close the game first, then run this again.")

    if revert:
        if st != "patched":
            print("The fix is not applied, nothing to undo.")
            return
        b = find_backup(exe, data)
        if not b:
            sys.exit("No matching backup found. In Steam use: right-click the game > "
                     "Properties >\nInstalled Files > Verify integrity of game files.")
        shutil.copy2(b, exe)
        print("Done: the original game file is restored.")
        return

    if st == "patched":
        print("The fix is already applied. Nothing to do.")
        return

    patched, info = build_patch(data)
    backup = exe + BACKUP_SUFFIX
    shutil.copy2(exe, backup)              # replaces a backup left from an older version
    tmp = exe + ".tmp"
    with open(tmp, "wb") as f:
        f.write(patched)
    shutil.copymode(backup, tmp)
    print("Backup of the original file:", backup)
    if "--verbose" in sys.argv:
        print("job offset %#x, cleanup site %#x, finish site %#x, cave %#x"
              % (info["x"], pe_va(info["site_a"]), pe_va(info["site_b"]), pe_va(info["cave"])))


def pe_va(rva):
    return BASE_DEFAULT + rva


if __name__ == "__main__":
    main()
SSnuffles20 2026-10-09 github

@p-valentin You are the real deal! I'm so impressed. I'm starting an electrical engineering degree in a year. The coding part scares me. I hope I can keep up with the studies. I will keeping using the script until there is an official update.

Thank you!

Mmarcel-pietron 2026-10-09 github
Slow save loads (90–120 s) on 5.00: XInput polling → setupapi rescan on every empty slot (cause + workarounds)

System: RTX 3060, NVIDIA 615.71, Fedora 44 (kernel 7.2), proton-cachyos 11.0-20261005 (same as Proton Experimental 11). Game 5.00c, DX12, --launcher-skip.

Symptom: loading any save, from the main menu or in-game, takes ~110 s. Only ~8 s of that is actual disk I/O and decompression. For the rest, MainThread and RenderThread spin while disk and GPU sit idle. Same symptom as #10231.

Cause: while loading, the game polls XInputGetState (imported from XINPUT9_1_0.dll) for all 4 slots in a tight loop: ~3,400 calls/s in a WINEDEBUG=+tid,+xinput,+setupapi log. In Wine's dlls/xinput1_3/main.c, xinput_get_state() calls update_controller_list() every time the requested slot has no device. That runs SetupDiGetClassDevsW(GUID_DEVINTERFACE_WINEXINPUT, DIGCF_PRESENT) plus interface enumeration, which is all registry access through wineserver. Every empty slot pays for a full rescan on every poll. eu-stack samples of the main thread during the stall were 100% in NtOpenKeyEx / NtQueryValueKey / enumerate_key / NtCreateKey / NtSetValueKey. Disconnected controllers left in the prefix (Xbox 045E:0B12 / 0B13, Steam virtual pad 28DE:11FF) make each scan more expensive.

Ruled out: mods / mod.io (-disablemods), network / VPN, disk, ntsync (active).

Workaround 1 (no controller connected): with the game closed, back up compatdata/292030/pfx/system.reg, then delete every [...] section whose key name contains your controllers' VID/PID: Enum\HID, Enum\WINEBUS, Enum\WINEXINPUT, Enum\USB and Control\DeviceClasses\{...}. Load time went from ~110 s to 19 s. Wine re-adds the entries the next time you connect a controller.

Workaround 2 (controller connected): put a small wrapper xinput9_1_0.dll next to witcher3.exe and launch with WINEDLLOVERRIDES="xinput9_1_0=n,b" %command%. It forwards to the builtin xinput1_3 but caches ERROR_DEVICE_NOT_CONNECTED per slot for 1 s, so empty slots stop triggering a rescan on every poll. The controller works normally and hotplug is picked up within ~1 s. Loads are fast again with the controller connected.

Wrapper source (build: zig cc -target x86_64-windows-gnu -O2 -shared -s -o xinput9_1_0.dll xiwrap.c xinput9_1_0.def)
/* xinput9_1_0 wrapper: caches "not connected" per slot for 1 s so Wine's
 * xinput doesn't rescan all devices (setupapi/registry) on every poll. */
#include <windows.h>

#define SLOTS 4
#define HOLD_MS 1000ULL
#define NOT_CONNECTED 1167 /* ERROR_DEVICE_NOT_CONNECTED */

typedef DWORD (WINAPI *get_state_fn)(DWORD, void *);
typedef DWORD (WINAPI *set_state_fn)(DWORD, void *);
typedef DWORD (WINAPI *get_caps_fn)(DWORD, DWORD, void *);

static get_state_fn real_get_state;
static set_state_fn real_set_state;
static get_caps_fn real_get_caps;
static volatile ULONGLONG empty_until[SLOTS];
static INIT_ONCE once = INIT_ONCE_STATIC_INIT;

static BOOL CALLBACK load_real(INIT_ONCE *o, void *p, void **c)
{
    HMODULE m = LoadLibraryW(L"xinput1_3.dll");
    if (m)
    {
        real_get_state = (get_state_fn)GetProcAddress(m, "XInputGetState");
        real_set_state = (set_state_fn)GetProcAddress(m, "XInputSetState");
        real_get_caps = (get_caps_fn)GetProcAddress(m, "XInputGetCapabilities");
    }
    return TRUE;
}

static BOOL cached_empty(DWORD i)
{
    InitOnceExecuteOnce(&once, load_real, NULL, NULL);
    return i < SLOTS && GetTickCount64() < empty_until[i];
}

static DWORD remember(DWORD i, DWORD ret)
{
    if (i < SLOTS) empty_until[i] = ret == NOT_CONNECTED ? GetTickCount64() + HOLD_MS : 0;
    return ret;
}

__declspec(dllexport) DWORD WINAPI XInputGetState(DWORD i, void *state)
{
    if (cached_empty(i)) return NOT_CONNECTED;
    if (!real_get_state) return NOT_CONNECTED;
    return remember(i, real_get_state(i, state));
}

__declspec(dllexport) DWORD WINAPI XInputGetCapabilities(DWORD i, DWORD flags, void *caps)
{
    if (cached_empty(i)) return NOT_CONNECTED;
    if (!real_get_caps) return NOT_CONNECTED;
    return remember(i, real_get_caps(i, flags, caps));
}

__declspec(dllexport) DWORD WINAPI XInputSetState(DWORD i, void *vib)
{
    if (cached_empty(i)) return NOT_CONNECTED;
    if (!real_set_state) return NOT_CONNECTED;
    return remember(i, real_set_state(i, vib));
}

xinput9_1_0.def:

LIBRARY xinput9_1_0
EXPORTS
    XInputGetState
    XInputSetState
    XInputGetCapabilities

Possible upstream fix: rate-limit update_controller_list() in xinput_get_state() / XInputGetCapabilities (e.g. at most once per N ms), since hid_update_thread_proc already picks up hotplug through device notifications and its periodic rescan.

Ggabrielmaialva33 2026-10-09 github

@p-valentin Heads-up: the new script doesn't actually patch the game. In main() it writes witcher3.exe.tmp but never moves it over the exe, so a normal run prints the backup line, exits 0 and leaves witcher3.exe untouched (--check still says NOT applied).

The rest checks out. On a copy of the 5.01 exe, the .tmp it leaves behind is byte-identical to what the 5.01 port produces, --check recognises a file patched by the port, and --revert restores it from the .orig-1048522 backup. It just needs the last step after shutil.copymode(backup, tmp):

    os.replace(tmp, exe)
    print("Done: the crash fix is applied. Start the game from Steam as usual.")

Also, the outer <summary> picked up part of the "What changed" list, so the collapsed header shows it twice.

Ggabrielmaialva33 2026-10-09 github

@marcel-pietron Nice find, and it matches Proton's source. The rescan on every empty slot came in with a4b10923 ("Rescan controllers before reporting disconnection to callers", cherry-picked from upstream Wine for RiME, CW-Bug #27003). It's on the experimental_11.0 branch, so Proton Experimental and the builds based on it (proton-cachyos, Wineland) have it. The proton_11.0 branch doesn't: there xinput_get_state() just returns not connected, so stable Proton 11.0 shouldn't have the slow loads. I haven't timed that, though.

Valve already ran into this with Far Cry 4 and added a hack that skips the rescan in xinput_get_state() for that game only: 1a1efb7f, CW-Bug #27566. Its commit message describes the same problem. Adding 292030 next to 298110 there would probably be the smallest fix on Proton's side.

Ccoreybruce 2026-10-10 github

Not to butt in the conversation but does anyone know why my game will just refuse to start with my previous reply?

Fflaax-io 2026-10-10 github

@p-valentin @gabrielmaialva33 some data from my side for 5.01 and the new script.

5.01 still has the race (crash address on the unpatched 5.01 exe): EXCEPTION_ACCESS_VIOLATION reading 0x0 at witcher3.exe+0x3302f3 in RenderDispatcher2, 131 s into the session, in the bestiary. That's the same instruction as +0x3317e3 in 5.00c (closure at +0x3302e0 in 5.01, offset +0x13). So the 5.01 port patches the right code.

5.01 with the fix: several hours of play (other testing, modding, etc) since 2026-10-08 without a crash (same setup as before: RX 9070 XT, proton-cachyos 11.0-20261005, mods, RT). Unpatched it crashed after 131 s.

v2 (pattern search), checked on copies of the original exes:

  • 5.00c and 5.01: build_patch() output is byte-identical to the v1 script (5.00c) and to Gabriel's 5.01 port (5.01). Same sites and cave (5.01: A 0x141bf7f4c, B 0x141bf189e, cave 0x1421e1258).
    • 5.00c fixed: 07b8efdb1ff4…, 5.01 fixed: 6313577b5689…
  • status() recognises "fix + the two wine_get_version RT bytes" as patched on both versions. Applying the fix to an exe that already has the RT bytes also works.
  • I can confirm the missing os.replace Gabriel found.

One more thing in v2: --revert after the RT patch. With the recommended order (fix first, then the sed/RT bytes), find_backup() returns None, because it only accepts a backup whose patched form equals the current file exactly, and the RT bytes differ. So --revert says "No matching backup found" although witcher3.exe.w3fix-original is right next to the exe. It fails safe, but people end up verifying the game files in Steam. A possible fix: accept the backup when build_patch(backup) and the current file differ only in a few bytes outside the 61 patched bytes, and print that those other changes (e.g. the RT bytes) are reverted too. Or tell people to undo the RT bytes before --revert.

(I hit the opposite problem in my own wrapper, so it's worth keeping v2's strict matching in spirit: after 5.01 replaced the exe, my first version would have copied the 5.00c backup over the new 5.01 exe on revert. v2's "backup must patch into the current file" check avoids that, and any looser match should keep that guarantee.)

Fflaax-io 2026-10-10 · hidden on GitHub github

@coreybruce I had a look at your log. The game does start, but it crashes about 1–2 s after witcher3.exe is loaded, before the window appears:

  • Main thread, right after Steam API and audio init: EXCEPTION_ACCESS_VIOLATION writing to address 0x0 at witcher3.exe+0x273a2b6 (5.01).
  • In 5.01 that instruction stores the first element of a list the game has just allocated. At the crash the element count is 0 (rdi=0) and the buffer pointer is NULL (rax=0). So some list the game builds at startup is empty, and the code doesn't expect that. Unfortunately the function has no strings, so I can't tell which list it is.
  • After that the game's crash handler gets stuck in dbghelp!MiniDumpWriteDump (fixme:dbghelp:MiniDumpWriteDump, then RtlpWaitForCriticalSection … blocked by 0144 every 60 s). That's why "nothing happens": the process stays alive without a window. I saw the same handler hang with older Wine builds. With proton-cachyos 11.0-20261005 and Proton Experimental the game's crash dialog appeared for me instead.

It is not the inventory crash.

Things that would help narrow it down:

  1. After a failed start, look in steamapps/compatdata/292030/pfx/drive_c/users/steamuser/AppData/Roaming/The Witcher 3/CrashInfo.json and …/AppData/Local/REDEngine/ReportQueue/. If a report folder appears, version.txt and the crash dialog's "Stacktrace" field (if you get the dialog with Proton Experimental) would show where it comes from.
  2. Your log is from Proton 11.0-2c with launch options --launcher-skip -fullscreen. Try once without -fullscreen.
  3. Did the game ever start on this machine with 5.00c, or never since the Remastered update?
KKaChino22 2026-10-10 github

Replying to https://github.com/ValveSoftware/Proton/issues/607#issuecomment-6077616400

Same issue as @coreybruce here: the game never shows a window after the Remastered update. I dug into it a bit, maybe this helps narrow it down.

System

  • CPU: Intel i7-8700K, GPU: AMD RX 6700 XT (RADV / Mesa 26.2.4)
  • OS: CachyOS, kernel 6.18.55-cachyos-lts
  • Game: GOG version 5.01, launched via Heroic 2.22.3 (with Comet 0.2.0) and directly via umu-run

Symptom
REDlauncher opens; after pressing Play (or with --launcher-skip) witcher3.exe keeps running but never creates a window (xwininfo -root -tree shows nothing). It sits at ~5 % CPU / ~1.4 GB RSS indefinitely. No crash, no error.

Tested, no change

  • Proton Experimental (11.0-20260910b), Proton 11.0 (stable), GE-Proton 11-5 / 11-6, proton-cachyos 11.0-20261005
  • Fresh prefixes for each Proton version
  • NVAPI disabled, PROTON_USE_PIPEWIRE=0, -disablemods, offline
  • No controllers connected

What the logs show (PROTON_LOG=1, WINEDEBUG=+loaddll,+debugstr,+winsock,+secur32)

  1. D3D12 device creation via vkd3d-proton succeeds (SM 6.8, DXR 1.1), Streamline initialises fine.
  2. Threads RedIO, DecompressionThread, UserProfileWorker start.
  3. Galaxy64.dll + bin\ddi\GOG.dll load. Without Comet the game retries 127.0.0.1:9977 ~19× for 15 s.
    With Comet running the handshake succeeds: Comet logs Client identified, sets up the gameplay DB and syncs leaderboards, no unhandled requests.
  4. Audio (Wwise / winepulse) initialises.
  5. After that the main thread goes silent. Nothing else is logged until I kill it.

So graphics, audio and the Galaxy connection all work, and the main thread then stalls before creating a window.

Launching bin/x64_dx12/witcher3.exe directly (without REDprelauncher) makes it exit cleanly ~0.3 s after start (AppPolicyGetProcessTerminationMethod), so it seems to require the prelauncher.

Proton log attached. Happy to test anything.

steam-0.log.gz

Follow-up: in my case it does not seem to be the same crash as @coreybruce.

  • My Proton Experimental log was recorded with +seh: no c0000005 / access violation and no MiniDumpWriteDump / RtlpWaitForCriticalSection in it.
  • AppData/Local/REDEngine/ReportQueue/ is created on first launch but stays empty in every prefix I tried, and CrashInfo.json (old prefix only) is empty (0 bytes).
  • I don't use -fullscreen.

So the game doesn't crash here. The main thread just stops after audio init, without any exception, and the process sits there without a window. I never got 5.00c / 5.01 to start; 4.04 worked fine before the update.

Kkisak-valve maintainer 2026-10-11 github

The Witcher 3: German QWERTZ keyboard layout is interpreted incorrectly

Issue transferred from https://github.com/ValveSoftware/Proton/issues/10242.
@paskoc posted on 2026-10-11T10:49:26:

Compatibility Report

  • Name of the game with compatibility issues: The Witcher 3: Wild Hunt GOTY / Remastered (GOG version)
  • Steam AppID of the game: 292030

System Information

  • GPU: NVIDIA GeForce RTX 3080 10 GB
  • Video driver version: NVIDIA 615.71.09
  • Kernel version: 7.0.0-38-generic
  • OS: Ubuntu 24.04.5 LTS
  • Desktop environment: GNOME on Wayland
  • Link to full system information report as Gist: https://gist.github.com/paskoc/32ff4b9e06a0dd48b7731047bdff7fd7
  • Proton version: Proton Experimental; also tested with Wineland 11.0-20261005

I confirm:

  • [ ] that I haven't found an existing compatibility report for this game.
  • [X ] that I have checked whether there are updates for my system available.

Proton Experimantal.log

steam-0 Wineland.log

Symptoms

On Ubuntu 24.04.5 LTS, The Witcher 3: Wild Hunt GOTY / Remastered appears to interpret my German QWERTZ keyboard layout incorrectly. The Z and Y keys are swapped in the game.

The issue occurs with both Proton Experimental and Wineland 11.0-20261005 when launching the GOG version through Heroic Games Launcher.

The Elder Scrolls Online works correctly with GE-Proton Latest on the same Ubuntu installation, without the Z/Y key swap.

I have checked the following:

  • Ubuntu uses the German keyboard layout correctly.
  • The Wine prefix registry contains keyboard layout ID 00000407 (German).
  • The game's keyboard_layouts_description.xml maps ID 00000407 to input_qwertz.
  • The game's input_qwertz.ini contains QWERTZ-specific bindings.
  • The saved input.settings contains entries such as IK_Y=(Action=CameraLock), consistent with the QWERTZ configuration.

The configuration appears correct, but the game still handles the keyboard input incorrectly. The exact cause is not yet known.

Reproduction

  1. Use Ubuntu 24.04.5 LTS with GNOME on Wayland and a German QWERTZ keyboard layout.
  2. Use the installed GOG version of The Witcher 3: Wild Hunt GOTY / Remastered.
  3. Launch the game through Heroic Games Launcher using Proton Experimental.
  4. Check the behavior of the physical Z and Y keys in the game.
  5. Repeat the test with Wineland 11.0-20261005.
  6. Observe that the Z and Y keys are interpreted incorrectly with both runners.

Separate Steam logs have been created for the tests with Proton Experimental and Wineland and can be attached to this report.

Nninele7 2026-10-11 github

Replying to https://github.com/ValveSoftware/Proton/issues/607#issuecomment-5899876406

Generated with LLM.

I profiled the stutter that appears when signed in to a CDPR account. While signed in, the game polls InternetGetConnectedState ~10 times per second on the main thread, and under Wine each call ends up re-reading /proc/net/dev once per network interface (via GetAdaptersAddresses -> nsiproxy.sys). Possibly related: Wine bug 60465.

Every read of /proc/net/dev queries the stats of all NICs, so a single slow driver is enough. On my machine (proton-cachyos 11.0-20261005, game 5.0.0.1048522) that is an Intel I226-V (igc): ~107 µs per stats read while its link is up, versus ~6 µs for the other interfaces. The number of interfaces (9 here) only multiplies how often that slow read happens, and it adds up to ~6 ms of blocked main thread per poll: one ~15 ms frame every 100 ms at an otherwise steady 7 ms.

Taking the igc link down (ip link set <if> down; disconnecting in NetworkManager is not enough) removed the stutter completely, same as signing out: p99 frame time 16 ms -> 8.8 ms.

Ggabrielmaialva33 2026-10-12 github

@ninele7 Wineland 20261011 (out yesterday) has a fix for exactly this. From its release notes: "Reduced stuttering in The Witcher 3 by avoiding repeated network adapter enumeration during frequent connectivity checks." Could be worth a try if you can switch from proton-cachyos: https://github.com/nanomatters/proton-cachyos/releases/tag/wineland-11.0-20261011

Ggabrielmaialva33 2026-10-12 github

@flaax-io @coreybruce @KaChino22 One more data point for the no-window crash. arrowgent hit the same thing on 5.00c (his log from Oct 4, Wineland 20260930): a write AV at witcher3.exe+0x2731cb6, mov [rax], r12 with rax=0, right after the call that allocates the list, and then the crash handler hanging. coreybruce's +0x273a2b6 on 5.01 is the same instruction: the bytes around it match, the code just moved by 0x8600. So it's not new in 5.01.

All three reports so far are on AMD (RX 7900M, RX 9070 XT, RX 6700 XT), but flaax-io's 9070 XT runs fine, so it isn't AMD alone. Something in those setups leaves that list empty at startup.

Ccrplcreek 2026-10-12 github

"Frozen" loading screen after dying

Under Proton, "Load last save" from the death screen can take 2–4 minutes, while the same save loads in seconds from the main menu (or from active session when selected from the options menu any time the player opts to load a save when they have not died). Reported on Steam Deck and desktop Linux, not on Windows; my testing and several reports showing Proton 11.0-02 cut it to 10-30 s Steam thread.

Quitting to main menu and loading from there is a possible workaround for effected versions until this is fixed.

System Information

  • GPU: AMD Radeon RX 9070 16 GB
  • Video driver version: 4.6 (Compatibility Profile) Mesa 26.2.4-arch3.1
  • Kernel version: 7.2.9-2-cachyos
  • OS: CachyOS (64 bit)
  • Desktop environment: KDE on Wayland
  • Link to full system information report as Gist:
    https://gist.github.com/crplcreek/32707c6d688f7c46c6884d1224f015f5
  • Proton version: Proton Experimental (1791480010 experimental-11.0-20261008-x86_64); also tested with Proton-CachyOS-Latest (1791146054 cachyos-11.0-20261005-slr)

EDIT: Possible this the same issue / related to another user's (much more detailed) comment above: link

Ggabrielmaialva33 2026-10-13 github

@crplcreek Yes, very likely the same thing as marcel-pietron's. The controller rescan that makes those loads slow came in with a Wine change that's only on Proton's experimental branch, so Experimental and proton-cachyos have it and stable Proton 11.0 doesn't, which fits what you saw with 11.0-02. Until Valve adds the game to the Far Cry 4 hack, his two workarounds should help: clear the disconnected controllers from the prefix, or the small xinput wrapper if you play with a controller. I'll update the gist row with your numbers and the cause.

Aalasky17 maintainer 2026-10-13 github

@gabrielmaialva33 (or anyone else) Unfortunately, we haven't been able to repro this slow save issue yet which makes it difficult to triage. If someone could attach a save file here that they are seeing the issue with, that could be helpful :) The save files should be located in ~/.local/share/Steam/steamapps/compatdata/292030/pfx/drive_c/users/steamuser/Documents/The Witcher 3/gamesaves -- if you just tar/zip up that whole gamesaves folder, that would probably be easiest.

Also - just to confirm since the issue is helped by disconnecting controllers - which controller(s) do you have plugged in when you are seeing this issue? Do you also have other peripherals plugged in (if so, which ones)? Also - if you are on Steam Client beta, could you try Steam Client stable and see if you can still reproduce the issue there?

Ggabrielmaialva33 2026-10-13 github

@alasky17 I haven't actually timed it on my machine, sorry. The numbers in this thread are marcel-pietron's (~110 s per load, 19 s after cleaning the prefix) and crplcreek's (2–4 min after dying). My part was matching it to the source.

My guess on why it doesn't repro for you: the per-poll rescan only exists on experimental_11.0 (a4b10923), so stable 11.0 won't show it at all. And how slow each rescan is depends on how many device entries the prefix has piled up. While loading, the game calls XInputGetState on all four slots a few thousand times a second, and each empty slot triggers a full SetupDi enumeration. A clean test prefix has almost nothing to enumerate. Prefixes that have seen controllers come and go (DualSense, Xbox 045E:0B12/0B13, the Steam Input virtual pad 28DE:11FF) keep those entries under Enum\HID / WINEBUS / WINEXINPUT / USB and Control\DeviceClasses. My Witcher 3 prefix, for example, has about 100 sections naming DualSense / Steam / Xbox IDs and 300+ DeviceClasses entries.

So I'd try: Proton Experimental, no controller plugged in, a prefix where a pad was connected and disconnected a few times, then load any save. It shouldn't depend on the save file; marcel sees it loading any save from the main menu too.

For your other questions: I use a DualSense, Steam client stable. If it helps, I can time loads here on Experimental vs 11.0 and attach my saves.

Proton versions

Launch options

Launch lines

Upstream links

DLLs

Error codes