World of Warcraft on mesa 20.3.4-1 will crash to black or corrupted screen
Description
When trying to play World of Warcraft through wine "lutris-6.0-rc1" build. The whole desktop will crash and the display will either stay black (no signal) or become corrupted. With the mesa version 20.3.4-1. This will happen within 5 minutes of gameplay. I could open a new tty, but to fully recover a reboot was needed.
World of Warcraft is set to run in DX12 mode.
I've had these kind of crashes for a while, but they used to be more rare. This is the second Sapphire Pulse 5700XT card that I'm using, the first one I sent back to the retailer and they sent me a new card. Both cards have seen these types of crashes.
System information
System: Host: arch-desktop Kernel: 5.10.14-arch1-1 x86_64 bits: 64 compiler: gcc v: 10.2.1 Desktop: KDE Plasma 5.20.5
tk: Qt 5.15.2 wm: kwin_x11 dm: SDDM Distro: Arch Linux
CPU: Info: 16-Core model: AMD Ryzen 9 5950X bits: 64 type: MT MCP arch: Zen 3 rev: 0 L2 cache: 8 MiB
flags: avx avx2 lm nx pae sse sse2 sse3 sse4_1 sse4_2 sse4a ssse3 svm bogomips: 217671
Speed: 3587 MHz min/max: 2200/3400 MHz boost: enabled Core speeds (MHz): 1: 3587 2: 2879 3: 2874 4: 2873 5: 2871
6: 2869 7: 2874 8: 3596 9: 3598 10: 2873 11: 2870 12: 2872 13: 2872 14: 2872 15: 2871 16: 2879 17: 2869 18: 2877
19: 2875 20: 2878 21: 2869 22: 2864 23: 2870 24: 2877 25: 2880 26: 3595 27: 2880 28: 2874 29: 2876 30: 2874
31: 2876 32: 2874
Graphics: Device-1: Advanced Micro Devices [AMD/ATI] Navi 10 [Radeon RX 5600 OEM/5600 XT / 5700/5700 XT]
vendor: Sapphire Limited driver: amdgpu v: kernel bus ID: 0c:00.0 chip ID: 1002:731f
Display: x11 server: X.org 1.20.10 compositor: kwin_x11 driver: loaded: amdgpu resolution: <missing: xdpyinfo>
OpenGL: renderer: AMD Radeon RX 5700 XT (NAVI10 DRM 3.40.0 5.10.14-arch1-1 LLVM 11.0.1) v: 4.6 Mesa 20.3.4
direct render: Yes
Log files (for system lockups / game freezes / crashes)
journalctl amdgpu-crashing-desktop.log
dmesg (separate crash from journalctl output) amdgpu.dmesg
Screen when it's corrupted and not just blacked out.
If applicable
- Xserver version: X.Org X Server 1.20.10
- DXVK version: v1.7.3L-03f11ba
- Wine/Proton version: lutris-6.0-rc1
Regression
Did it used to work in a previous Mesa version? It can greatly help to know when the issue started.
- Downgrading to use mesa 20.3.1 or 20.3.2 instead will make the issue occur less frequently. It would still happen every other day to not seeing it for over a week, now it instead happens consistently 5-10 minutes into gameplay.
API captures (if applicable, optional)
Consider recording a GFXReconstruct (preferred), RenderDoc, or apitrace capture of the issue with the RADV driver active. This can tremendously help when debugging issues, but you're still encouraged to report issues if you can't provide a capture file.
- I did a capture with GFXReconstruct, though the file ended up at 68GB and it looks to take 10 hours if I want to upload it to google cloud to share. Funnily enough trying to replay the file I crashed within 10-20 seconds. I'll see if I can get a capture that isn't quite as big.
Further information (optional)
Does the issue reproduce with the LLVM backend (RADV_DEBUG=llvm
) or on the AMDGPU-PRO drivers?
Does your environment set any of the variables ACO_DEBUG
, RADV_DEBUG
, and RADV_PERFTEST
?
- None of these variables should be set, I did try with
RADV_DEBUG=llvm
, still crashing the same way as far as I can tell. - Running with
RADV_DEBUG=hang
did not generate any folder named$HOME/radv_dumps_*/