Frequent crashes / hangs with message "GPU has fallen off the bus"

Hello, I am running Gentoo Linux with latest stable nvidia binary drivers with a GTX 3080 Ti. Since I installed the card, unfortunately my computer hangs very often (2/3 times per day) which forces a hard reboot. I’m certain this is due to the nvidia card, as I used the internal Intel GPU before for 6 monthes without any issues.

This almost always happens when using Firefox / WebGL, but I don’t do much gaming on this box so this might be because this is the only program using 3D / the nvidia card.

When this happens, display gets completely frozen, mouse locks, but I still can ssh into the machine from another box. I ran the bug report from such a frozen situation. I attach it as well as the dmesg log which contains a lot of errors related to nvidia at the end. Seems like the important message is [45044.053884] NVRM: Xid (PCI:0000:01:00): 79, pid=‘’, name=, GPU has fallen off the bus.

I hope a nVidia engineer will be able to help / trace this issue. I’ve googled a bit and suggestions are that this may be a hardware / power supply issue. However, in this case it would probably also happens under Windows, and I never noticed it (although tbh I don’t use Windows much on this computer). Also, this does not happen when playing intensive games at all… only desktop work, with as I said very limited WebGL games on some tabs.

This is very, very frustrating having your main computer hang twice or three times a day like that when everything else is fine, so I hope someone can help.

nvidia-bug-report.log.gz (848.5 KB)
dmesg-nvidia (76.5 KB)

Hello, up? Can I get any advice or recommendation?

I have a similar issue, but with 3090 founders edition, ubuntu 20.04, and drivers 535.104.12

What PSU do you use? Mine is 600W. Could it be not enough for the 3080 Ti?

I use a Super Flower 1000W 80 Plus Platinum

Well then I doubt it is a PSU issue, at least in your case. I dont think it is either for me because this never happens under Windows.

Bump?

Very unclear why the gpu is falling off the bus by just doing normal desktop tasks. To check whether there might be some subtle psu issue involved, please limit gpu clocks
nvidia-smi -lgc 300,1500
then check if the gpu is stable.