protonscr

Assassin's Creed Origins; DXVK-2.6.0 last good version (Win11 + AMD)

dxvkopen can't reproducewindowsamd proprietary
doitsujin/dxvk#5813 · opened 2026-07-28 by chaosophy · updated 2026-07-30 · 9 comments · github
2 matching comments, n / p to jump
Cchaosophy 2026-07-28 github

Software information

Assassin's Creed Origins (Game ver 1.62)
Max settings, no DRS, no HDR.

System information

  • OS: Win 11 Pro 25H2
  • GPU: AMD 9070XT
  • Driver: 26.6.1
  • Wine version: n/a
  • DXVK version: 2.6.0 up to 3.0.2

Apitrace file(s)

  • Have tested eight versions, so haven't got around to doing any traces but happy to do some if/when needed on specific versions.

Log files

  • As above. I wasn't saving log files in testing but happy to re-run and send as needed.

Basic description of issue

3.0.2/3.0.1/3.0.0

  • All exhibit an issue with shadows.
  • Had some crashes/driver timeouts with I think 3.0.2 running the benchmark and then in game.
  • 3.0.0 has issues on the map screen but 3.0.1/2 seem to be ok.

2.7.1/2.7.0

  • The game initially opens ok but will insta crash trying to load a game or the benchmark.

2.6.2/2.6.1

  • These don't work too bad but have issues with the map screen.

2.6.0

  • This version seems to be perfect. I have played multiple hours now and not noticed any issues, out side of what just seem to be inherent to the game. TBH I've only played a little bit of the game without DXVK, so not doing any pixel peeping before/after comparisons but all seems ok.

Additional

A few (all?) versions after 2.6.0 also exhibit issues on the title screen, I was trying to make notes of which versions but tbh it's kind of inconsistent. As far as I can tell 2.6.0 is fine on the title screen.

My scores in the in game benchmark go from ~12k to ~18.5k so really happy with the performance uplift.

The game comes with 'amd_ags_x64.dll' version 5.1.1.0 and appears to be replaceable. Not really seen much difference but now that I've got the game running with DXVK 2.6.0 I've swapped it out with version 6.3.0.0 from 'FidelityFX-Samples-v2.3.0'. All above testing was done with the game's version. Probably pointless swapping it out and I'm not really seeing a difference but it's there for people who like to mess with these things.

Apologies for not uploading log files etc but it took me a few days of trying different versions and now I have one that works I've just been enjoying the game. :D

Let me know what you need from me and I'll try and supply more info.

Thanks!

Ddoitsujin maintainer 2026-07-28 github

My crystall ball says that you're likely getting that Error -13 thing on swapchain creation, which is a common issue on some AMD Windows setups and isn't really fixable. Try messing around with Windows HDR settings or something.

Kinda impossible to know for sure w/o log file, but yeah, game works fine on Proton, Windows is just a complete mess.

All exhibit an issue with shadows.

Not sure what this means?

The game comes with 'amd_ags_x64.dll' version 5.1.1.0 and appears to be replaceable.

Also, don't do that. Different AGS versions are not ABI-compatible.

Cchaosophy 2026-07-28 github

Thanks for the quick reply.

HDR is disabled in Windows, in game and on the monitor.
Monitor does 'do' HDR but it's not great, haven't tried it with AC:O.

Shadows are visibly broken.
https://ibb.co/m5b78g2h

And yeah the DLL replace is just me messing about and tbh doesn't seem to make a difference, played a few hours last night with no issues, will probs swap back to the old one before my next sesh.

Game works fine with DXVK-2.6, so happy to just continue using that version.
At least until another driver update breaks it and I go through all this again. :D
Though to be fair the game doesn't run that bad with out DXVK, just that it runs so much better with it.

Ddoitsujin maintainer 2026-07-28 github

FWIW, I just tested the proprietary driver on my 6900XT (yes, older driver obviously because this is RDNA2) and even that works completely fine, including shadows. No idea what's happening there but once again likely to be a driver quirk.

Game works fine with DXVK-2.6, so happy to just continue using that version.

Just annoying that this is necessary because 2.6 is going to be quite a bit slower than 2.7 or 3.0 in this game.

Cchaosophy 2026-07-28 github

Yeah, it'd be nice to just be able to use the latest version but it's a nine year old game at this point, so just happy it works at all.

It spends most of it's time over 100 fps and regularly hits my fps cap (180Hz monitor), though I've been slowly dropping the fps cap as the game seems to have the usual physics and animation errors (mainly in cut scenes) at high fps so trying to see if lowering it helps.
Not sure if this is a base game issue or dxvk but it's rare and not really an issue for me.

If I'm still playing the game when a new driver*/dxvk version releases I'll try and remember this thread and re-test.

Thanks.

edit: * Driver 26.7.1 came out today.
Tried DXVK 3.0.2, same shadow issue.
Gone back to using 2.6.0

DDigger1955 2026-07-29 github

Greetings, @chaosophy

I am DXVK-GPLALL developer. Please, can you check DXVK-GPLALL 2.6.8-2? I am interested in your feedback.

Cchaosophy 2026-07-30 github

Hi @Digger1955 ,

I tried 'DXVK-GPLALL-MSVC-Windows-AVX2-O1_Oi_Ob3-LTCG-AMD64.2.6.8-2.zip', think this is the correct version for me running Win11, Ryzen 3900X, AMD 9070XT?

It seems to work ok for me, though it has the title screen glitch and what looks like a small glitch in the map screen, which aren't there in DXVK-2.6.
GPLALL also gives slightly lower scores in the in built benchmark, compared to standard DXVK but not by a lot.

DDigger1955 2026-07-30 github

@chaosophy

Thanks for the report.

I tried 'DXVK-GPLALL-MSVC-Windows-AVX2-O1_Oi_Ob3-LTCG-AMD64.2.6.8-2.zip', think this is the correct version for me running Win11, Ryzen 3900X, AMD 9070XT?

Yes, that one is correct.

Can you provide additional info?

  1. DXVK log;
  2. Screenshot of the Title Screen Glitch;
  3. Screenshot of the Map Screen Glitch;
  4. Please specify vs. which versions of standard DXVK GPLALL 2.6.8-2 has lower scores in the in built benchmark.
BBlisto91 2026-07-30 github

It would be better to move the back forth of those problems over to your issue tracker

Cchaosophy 2026-07-30 github

It would be better to move the back forth of those problems over to your issue tracker

No probs, have started a new thread over there.
https://github.com/Digger1955/dxvk-gplall/issues/109

DLLs