My Ryzen 5 5600GT and RX 5600 XT crash under GPU-heavy use with one kernel, but work with the regular Linux kernel. The logs mention the GPU disappearing from the bus, and after a crash it sometimes is not detected until I wait or reseat it. I also tried several desktop environments, with no change. What would you check next?
Why would this GPU crash only on one kernel?
Asked by
silted_lens
Answers
The fact that changing kernels changes the result is a useful clue, but it does not prove the kernel is the only problem. Try another available kernel variant, such as the long-term support build, and keep the working kernel installed so you can boot back into it. If that variant is stable, it points toward a compatibility difference between kernel builds rather than the desktop environment.
Reply to this answer
Those messages sound like the GPU stopped responding over the bus, followed by graphics-driver errors. The scheduler error may be a consequence of that loss rather than the original cause. Since the card can disappear from the firmware screen after a crash, I would also check temperatures, power connections, and whether the card is firmly seated. A working kernel is a reasonable temporary workaround, not a reason to assume the hardware is definitely fine.
Reply to this answer
Since several desktop environments behaved the same way, I would stop spending time swapping desktops. Compare the kernel versions and configuration of the working and failing kernels, and change one thing at a time. If only the specialized kernel crashes under load, staying on the working one is a sensible choice unless you specifically need the other build’s features.
Reply to this answer
Add your answer
Answers are reviewed before publication.
silted_lens:
I had not tried the other kernel variants yet. I’ll keep the regular kernel available while testing them.