Hi, recently my nvidia kernel driver is getting stuck, journalctl tell GPU dropped off the bus, screen goes black, fans spinning to highest. Even GDM restart won´t be able to use graphics again. Reboot from SSH is possible.
The same computer, until recently have been quite rock-stable.
Nvidia version 545.29.06 from rpmfusion-nonfree-nvidia-driver, NVIDIA GeForce RTX 3060, Fedora 39
Kernel 6.7.3-200.fc39.x86_64
nvidia-bug-report.log.gz (301.1 KB)
Display crashes by 15 minutes when running Steam game.
The gpu shut off, I suspect due to lack of power. Please check your psu and cables.
I have inspected hardware, everything seems fine. I took off the GPU from the slot and inserted it back. Reattached all power cables. Hopefully that might solve potential hard to find issues. I dont have spare PSU, but no other indication besides freezing GPU, I might obtain another if issue persists.
You might limit gpu clocks, e.g. running
nvidia-smi -lgc 300,1200
and check if it then runs stable. This would point to the psu.
Actually, it seems that re-inserting GPU and re-attaching all power cables indeed resolved stability issues! Since then, I have my system stable again. I have also reverted to older kernel and updated to newest kernel 6.7.4 - both were stable.
Kernel 6.7.3 was bugged, preventing nvidia driver to be compiled. Urgent 6.7.4 fixes this, but it very possible that the driver version from rpmfusion could meet issue because of this.
So, the probability that your hardware is in cause is very low.
I can really tell about rpmfusion drivers, I’m using my own tool for installation allowing me to fast switch back in case of problems. And I use 535.154.05 version.
You can try nvidia-persistenced service to force power on the gpu at boot. And see if it change something.