Desktop plasmashell dumpig core while gaming. New Radeon 9070xt .
Hey folks, it's me again (my previous post: https://discord.com/channels/1072614816579063828/1432852153063706624). The core dump happend again. This time I'm on pure Bazzite no layers, with Radeon 9070xt. Attached journal logs for dumps. Both plasmashell then Xwayland, then other app dumped cores as well, but i guess it's not that important. The crash results in crashing and closing couple of apps, but recovered into shell (no logout this time). The crashes are really random, had no issue for the whole day yesterday and today it crashed like 20 minutes into the game.
I really hope it's not the card itself - any tips on checking whether it's not a hardware issue?
Solution:
Jump to solution
yep, just finished 4 hours gaming session, no crash anymore, I've limited the power limit to 300W and downclocked the GPU clock by 425, LACT looks like this:
18 Replies
I've noticed a weird pattern. Looks like the crashes happen when i'm playing a game with Vesktop on which has hardware acceleration turned on. When I've turned the hardware acc off i'm unable to reproduce the crash - so far. But when it's on I've crashed twice in about 20 minutes after starting a game. Vesktop is also one of few application that survives the crash - most of others are core dump[ing. Would that ba any hint? Is is possible that Vesktop hardware acc is causing all of this?
what does the journald log look like?
We need that log, a coredump on its own is not very helpful. anything can cause a coredump chain that isn't plasmashell.
here are logs from ujust logs-last-boot
crashes happen at 2:24 and 3:17
its not plasma
its amdgpu
Nov 01 02:45:40 kernel: amdgpu 0000:03:00.0: amdgpu: Dumping IP State
Nov 01 02:45:40 kernel: amdgpu 0000:03:00.0: amdgpu: Dumping IP State Completed
Nov 01 02:45:40 kernel: amdgpu 0000:03:00.0: amdgpu: [drm] AMDGPU device coredump file has been created
Nov 01 02:45:40 kernel: amdgpu 0000:03:00.0: amdgpu: [drm] Check your /sys/class/drm/card1/device/devcoredump/data
Nov 01 02:45:40 kernel: amdgpu 0000:03:00.0: amdgpu: ring gfx_0.0.0 timeout, signaled seq=29524606, emitted seq=29524608
Nov 01 02:45:40 kernel: amdgpu 0000:03:00.0: amdgpu: Process GameThread pid 39132 thread vkd3d_queue pid 39195
Nov 01 02:45:40 kernel: amdgpu 0000:03:00.0: amdgpu: Starting gfx_0.0.0 ring reset
do you have any undervolt or overclock?
nope, turned them off couple of days ago
also just finished 3 hour long gaming session with discord hw acc turned off and brave (with hw acc on) working in background - no issues ... this is utmost puzzling
Does it only happen in WoW?
nope, WOW (very seldom in the past) and Expedition 33 (more often)
i can try other games as well if that'll make investgation easier
Not that this helps, but I'm also having the same issue. Some games work without issues (eg PEAK), while others immediately blackscreen and I have to reset my pc (OSRS, Hunt Showdown, etc). My logs have the same amdgpu entries - RX 5700 here no OC or UV. Toggling Vesktop HW Accel didn't do anything for me.
yep, got a crash with vesktop hardware acc off jus ta minute ago ...
Looks like i'm not the only one with this issue on Radeo 9070xt... I'm going to try tinker with limmiting the GPU power draw using LACT following this post: https://www.reddit.com/r/linux_gaming/comments/1lnhlxl/the_9070xt_is_making_it_really_hard_not_to_go/
Reddit
From the linux_gaming community on Reddit
Explore this post and more from the linux_gaming community
Solution
yep, just finished 4 hours gaming session, no crash anymore, I've limited the power limit to 300W and downclocked the GPU clock by 425, LACT looks like this:
I assume that current driver is puting to much load on the card and that causes it to crash. Or i'm super unlucky with the silicone and my card is actualy sub-par. Anyway i don't think it has anything to do with bazzite.
@pure - why don't you try downcloking yours and see if that solves it for you
I'll have to give that a shot and report back. Different card so I'll play with it for a bit. Thanks for the heads up!
I've deep dove into specs and defaults for my card (Radeon Sapphire 9070xt Pulse) - here are reference specs: https://www.techpowerup.com/gpu-specs/sapphire-pulse-rx-9070-xt.b12275
Looks like the card should not boost over 2970 MHz, but with out-of-the-box config and no overclock it was boosting to 3350 MHz, reaching power throttle state. This i believe was causing it to crash. With -450 offset to GPU clock the card sits within the product specs. Not sure where LACT and/or drivers take the reference values from, but it was clearly to high. I'll leave out the -450 offset and see if I will manage to crash (hopefully not 🤞 ) and report back in couple of days if i haven't (and close the thread) or keep the investigation going. Thanks again for all the support!
Yeah my clocks and power limit were definitely a bit off so I corrected those but to no avail. I appreciate the help though! I'll split my issue out into a separate thread as to not keep yours going.
5 hour or Epedition 33, no crash, Vesktop with hardware acceleration on (this was probably stupid and wrong lead), I guess you just need to under-clock the card if it's crashing. Gonna close the ticket.