AMD RX 6800 keeps crashing when playing Diablo 4 via Steam with Mesa 24.1.x installed

[jesus96]

Hello,

Hoping I can get some help here. I was inclined to think this is a hardware issue until my GPU didn't crash on Kubuntu which is using Mesa 24.0.9 but all other distros I have tried (currently on Debian Testing) use some version of 24.1 and all of these distros keep crashing when I play Diablo 4 on Steam. I can't speak much on other games as this is currently the game I've been playing but I can definitely test out some other games later.

The issue is as follows:
I will start playing and anywhere between 15 - 20 minutes into playing I get a black screen, fans kick up to 100% (sometimes they don't and they remain at their normal speed) and the last few seconds of audio will loop and sound distorted (sometimes no audio will play) and eventually my monitor displays the signal lost message. This happens every single time and the only way to be able to use my system again is to fully shut down and turn back on; the reboot button on my PC case won't do anything. The only distro this hasn't happened on is Kubuntu and it's the only one not running 24.1; i also tested on my Windows partition and it runs perfectly there as well.

Fairly certain I've already ruled out other PC components as being the issue especially since it previously had a 3080Ti that I was using temporarily and the system was running just fine. I grabbed the logs from journalctl and isolated what appears to be from the time I started playing to the last message that was logged before I had to shut down. I don't see anything that could hint at an issue and the last bunch of messages appear to be uneventful but I was wondering if maybe someone else would be able to find something that maybe I'm overlooking. I've attached the output logs to this post.

[CwF]

the last few seconds of audio will loop and sound distorted (sometimes no audio will play) and eventually my monitor displays the signal lost message.

I've seen this with amd gpu's and it is frustrating. Sometimes clues are missing from the end of the logs, sometimes the logging is active but will not be written to disk. Can you ssh in from another computer? Logged into a user (same as local is ok) with journalctl rights, have a 'journalctl -f' terminal active on the other computer and work it through the crash.

I take the audio loop as evidence part of the system is alive, and you may find some entries useful for a bug report narrowed down to the fail, such as'

Code: Select all

[drm:atom_op_jump []] *ERROR* atombios stuck in loop for more than 5secs aborting
[drm:atom_execute_table_locked []] *ERROR* atombios stuck executing CF7C (len 62, WS 0, PS 0) @ 0xCF98
0000:03:00.0: ring 3 stalled for more than 385956msec
0000:03:00.0: GPU lockup (current fence id 0x000000001afef942 last fence id 0x000000001afef943 on ring 3)

In addition, if you are familiar with x2x it may simultaneously report as logging continues, if using X;

Code: Select all

XIO:  fatal IO error 11 (Resource temporarily unavailable) on X server ":0.0"
      after 1590114 requests (1590114 known processed) with 0 events remaining.

It's been awhile, but as I remember I've had remote logging possible past the 'no signal' mark. If the power button is setup to shutdown it becomes unable to very early in the sequence - sometimes if pressed during the first few audio loops - usually not.

[D2b2426R5d]

I have RX 6800XT card and Diablo IV plays very well on it. I use Battle.Net (not Steam) client, and Debian Stable, with all default Stable versions of drivers and Mesa. I never had any crash when playing this game.
Maybe try Debian Stable and see how it goes. And as CwF, suggested, log in via ssh from another computer, and have logging open, another command to try is:

Code: Select all

sudo dmesg -w

That will show you all messages even after GPU has crashed.