protonscr

[Intel] DXVK 2.5 - 2.5.2 with DXVK_HUD=compiler freezes the game or the entire system as soon as "Compiling Shaders..." appears on the screen (Works without compiler HUD))

dxvkclosed intel anv
doitsujin/dxvk#4579 · opened 2025-01-04 by mtjsg · updated 2025-01-10 · 10 comments · github
1 matching comments, n / p to jump
Mmtjsg 2025-01-04 github

So, I have been having a problem with DXVK freezing games or the entire system since v2.5, and today is discovered that it's compiler option for DXVK_HUD that it's causing it.

Any game using it will freeze as soon "Compiling Shaders..." appears on screen, and sometimes my entire system (Fedora 41) freezes.

EDIT:
Forgot to mention, that when the system freezes, the keyboard numlock/capslock,scrollock keeps responding, audio continues to play, but I cannot switch to a tty.

Software information

Any game , any wine version, Bottles and Heroic launcher (all flatpaks)

System information

  • OS : Fedora 41
  • GPU: Intel CoffeeLake-S GT2 [UHD Graphics 630]
  • Driver: Mesa 24.3.1 (git-c815d651b8) i915
  • Wine version: Proton-GE 9.22, soda 9.0, Proton-GE 9.15, kron4ek-wine-10.0-rc2-staging-tkg-amd64 (same behavior on all versions)
  • DXVK version: 2.5 - 2.5.2 (2.4.1 its the last version to work)

Logs:

https://0x0.st/8iqL.log

Ddoitsujin maintainer 2025-01-04 github

All this does is render some text on screen in exactly the same way as the rest of the HUD, I don't really see a universe where this could cause the entire system to freeze.

Mmtjsg 2025-01-04 github

All this does is render some text on screen in exactly the same way as the rest of the HUD, I don't really see a universe where this could cause the entire system to freeze.

I was surprised too, I will try to record with my phone

Mmtjsg 2025-01-04 github

All this does is render some text on screen in exactly the same way as the rest of the HUD, I don't really see a universe where this could cause the entire system to freeze.

video:

https://photos.app.goo.gl/Pv6xoqNupDs4UDS97

EDIT: I can try to get the system freeze recording if you want to. But basically, removing the compiler hud option makes everything works ok.

BBlisto91 2025-01-04 github

I am not reproducing it with my Arc b580 at least. And that is also using Heroic to launch a game with same Proton and driver version. Though i believe that driver might not be the exact same what a older iGPU is using.

Mmtjsg 2025-01-05 github

it seems that DXVK_HUD=fps have the same behavior and noticied these errors when a game freezes with dmesg:

21:45:31 kernel: i915 0000:00:02.0: [drm] GPU HANG: ecode 9:1:eff7fffa, in VampireSurvivor [81043]
21:45:31 kernel: i915 0000:00:02.0: [drm] GPU HANG: ecode 9:1:eff7fffa, in VampireSurvivor [81043]
21:45:31 kernel: i915 0000:00:02.0: [drm] VampireSurvivor[81043] context reset due to GPU hang
21:45:31 kernel: i915 0000:00:02.0: [drm] GPU HANG: ecode 9:1:f1973ffc, in VampireSurvivor [81043]
21:45:31 kernel: i915 0000:00:02.0: [drm] GPU HANG: ecode 9:1:f1973ffc, in VampireSurvivor [81043]
21:45:31 kernel: i915 0000:00:02.0: [drm] GPU HANG: ecode 9:1:f1973ffc, in VampireSurvivor [81043]
21:44:37 kernel: i915 0000:00:02.0: [drm] GPU HANG: ecode 9:1:86dffffe, in VampireSurvivor [80192]
21:44:37 kernel: i915 0000:00:02.0: [drm] VampireSurvivor[80192] context reset due to GPU hang
21:44:37 kernel: i915 0000:00:02.0: [drm] GPU HANG: ecode 9:1:f1973ffc, in VampireSurvivor [80192]
21:44:37 kernel: i915 0000:00:02.0: [drm] GPU HANG: ecode 9:1:f1973ffc, in VampireSurvivor [80192]
21:44:37 kernel: i915 0000:00:02.0: [drm] GPU HANG: ecode 9:1:f1973ffc, in VampireSurvivor [80192]
21:44:37 kernel: i915 0000:00:02.0: [drm] GPU HANG: ecode 9:1:f1973ffc, in VampireSurvivor [80192]
21:44:37 kernel: i915 0000:00:02.0: [drm] GPU HANG: ecode 9:1:f1973ffc, in VampireSurvivor [80192]
21:44:36 kernel: i915 0000:00:02.0: [drm] GPU HANG: ecode 9:1:f1973ffc, in VampireSurvivor [80192]
21:30:11 ghostty: info(grid): reallocating GPU buffer old=2037 new=3063
21:15:03 resources: INFO resources::utils::gpu > Found GPU "CoffeeLake-S GT2 [UHD Graphics 630]" (PCI slot: 0000:00:02.0 · PCI ID: 8086:3e91 · Category: Intel)
20:57:08 ghostty: info(grid): reallocating GPU buffer old=38 new=65
12:47:24 resources: INFO resources::utils::gpu > Found GPU "CoffeeLake-S GT2 [UHD Graphics 630]" (PCI slot: 0000:00:02.0 · PCI ID: 8086:3e91 · Category: Intel)
12:40:12 ghostty: info(grid): reallocating GPU buffer old=2037 new=4602
12:34:28 resources: INFO resources::utils::gpu > Found GPU "CoffeeLake-S GT2 [UHD Graphics 630]" (PCI slot: 0000:00:02.0 · PCI ID: 8086:3e91 · Category: Intel)
11:23:41 ghostty: info(grid): reallocating GPU buffer old=897 new=1353
10:00:18 kernel: i915 0000:00:02.0: [drm] GPU HANG: ecode 9:1:8ed1fff2, in Backend [15568]
09:34:18 resources: INFO resources::utils::gpu > Found GPU "CoffeeLake-S GT2 [UHD Graphics 630]" (PCI slot: 0000:00:02.0 · PCI ID: 8086:3e91 · Category: Intel)
09:33:22 kernel: i915 0000:00:02.0: [drm] GPU HANG: ecode 9:1:8ed1fff2, in Backend [12643]
09:21:54 ghostty: info(grid): reallocating GPU buffer old=255 new=390
09:19:09 heroic: GPU 0:
09:18:47 ghostty: info(grid): reallocating GPU buffer old=897 new=1353
09:07:53 kernel: i915 0000:00:02.0: [drm] GPU HANG: ecode 9:1:8ed1eff2, in Backend [8419]
09:07:00 resources: INFO resources::utils::gpu > Found GPU "CoffeeLake-S GT2 [UHD Graphics 630]" (PCI slot: 0000:00:02.0 · PCI ID: 8086:3e91 · Category: Intel)
09:06:35 kernel: i915 0000:00:02.0: [drm] GPU HANG: ecode 9:1:8ed1fff2, in Backend [7920]
09:04:54 resources: INFO resources::utils::gpu > Found GPU "CoffeeLake-S GT2 [UHD Graphics 630]" (PCI slot: 0000:00:02.0 · PCI ID: 8086:3e91 · Category: Intel)
08:55:43 kernel: i915 0000:00:02.0: [drm] GPU HANG: ecode 9:1:8ed17ff2, in Backend [5932]
08:55:00 resources: INFO resources::utils::gpu > Found GPU "CoffeeLake-S GT2 [UHD Graphics 630]" (PCI slot: 0000:00:02.0 · PCI ID: 8086:3e91 · Category: Intel)
08:27:05 gnome-shell: Boot VGA GPU /dev/dri/card1 selected as primary

BBlisto91 2025-01-05 github

I think it is probably best to make a mesa bug report for this

Mmtjsg 2025-01-06 github

Lionel Landwerlin
@llandwerlin Replied with this in the mesa bug report:

Running with INTEL_DEBUG=sync (waiting on each submission) there is no hang.
Sounds like a DXVK issue, likely not synchronizing properly it's rendering with the application or something.
Can you file a bug over there?

And closed the bug with "Not Our Bug"

Mmtjsg 2025-01-06 github

Captura de tela de 2025-01-06 17-04-30

I still getting freezes and gpu hangs in dmesg with INTEL_DEBUG=sync.

here is a full dmesg:
dmesg.log

Launch options