6700 XT: Failed to export SMU metrics table (Linux 6.7)

Before submitting your bug report:

Brief summary of the problem:

I am getting random GPU crashes with Linux 6.7. I've always had the odd crash or two but previously the GPU Reset / Recovery handled this correctly. With 6.7 they seem more frequent and all displays turn off with no recovery.

Hardware description:

• CPU: AMD Ryzen 5950X
• GPU: 0d:00.0 VGA compatible controller [0300]: Advanced Micro Devices, Inc. [AMD/ATI] Navi 22 [Radeon RX 6700/6700 XT/6750 XT / 6800M/6850M XT] [1002:73df] (rev c1)
• System Memory: 64GB
• Display(s): 32" DELL G3223Q Connected over Display Port, LG Ultrawide (2560x1080) connected over HDMI

System information:

• Operating System: Gentoo 2.14
• KDE Plasma Version: 6.0.80
• KDE Frameworks Version: 6.0.0
• Qt Version: 6.6.1
• Kernel Version: 6.7.4-zen1 (64-bit)
• Graphics Platform: Wayland
• Processors: 32 × AMD Ryzen 9 5950X 16-Core Processor
• Memory: 62.7 GiB of RAM
• Graphics Processor: AMD Radeon RX 6700 XT

How to reproduce the issue:

I can't reliably reproduce it but the issues happen persistently after some random amount of time or random amount of CPU/GPU load, etc

Attached files

kern.log

It just crashed again after being on all day, this time the display stayed on.

Right after it crashed I see this in dmesg:

[Sun Feb 11 14:30:26 2024] [drm:amdgpu_dm_plane_helper_prepare_fb [amdgpu]] No FB bound
[Sun Feb 11 14:30:26 2024] [drm:amdgpu_dm_atomic_commit_tail [amdgpu]] VRR packet update: crtc=91 enabled=0 state=2

I wonder if VRR is the issue here? Right now I have it on automatic in Kwin's settings, I'll switch it to Never and see if that improves stability. kern.log

Quick update, my systems been running all day now with VRR disabled and in that time I've only had one single crash which triggered a graphics reset which Kwin successfully handled and restarted everything.

Perhaps my issue is related to #3149 (closed)

I do also have resizable BAR enabled but don't think that's related as I've been running with that turned on for ages now with no significant issues.

added 6000 dGPU Series label

added DC label

I'm closing this now.

Recent kernels contained a lot of logs like this in dmesg: amdgpu 0000:0c:00.0: amdgpu: device lost from bus!

This is odd to say the least, the device should not be disappearing from the bus. I took the system a part and re-seated the graphics card and have had no issues since. It's been running stable all week. I guess computers have ghosts sometimes.

closed