Potential problem with nVidia driver 531.18

[jonault]

Has anyone else been having problems after upgrading to the latest nVidia drivers on Windows 11 (531.18)? My GPUs are constantly dumping WUs now, twice this evening my computer spontaneously rebooted & once I had to manually reboot it because both GPU folding cores failed to restart after core shutdowns.

I'm going to try downgrading back to the previous driver to see if that fixes anything. I don't think the WUs are the problem, the project numbers are all over the place & not part of the same series.

(I should note this machine is running the beta 8.1.14 software, but it seemed to be working fine until the driver update.)

[calxalot]

There may be a problem with GPUs and v8.1.14.
You might want to reinstall 8.1.13.

[jonault]

It's been almost an hour and a half (edit: 2 hours) since downgrading the driver & there have been no problems. I'll check again in the morning but so far it looks like the issue was with the driver & not with 8.1.14.

[jonault]

It looks like I might have jumped the gun here. When I had both GPU cores stop & not restart last night I assumed that was related to all the dumped work units. But when I checked the logs this morning I found that I had dumped 6 more WUs overnight but that they were all CPU work units - all GPU work units completed normally. I then went back & checked the logs from yesterday evening and all the dumped WUs there were CPU as well. And I had another spontaneous system reboot while examining the logs.

I don't know if the GPU cores stopping last night had anything to do with the video driver or not, but all those dumped work units are not related. For now I'm stopping CPU folding on that machine until I can figure out what the problem is. The Windows system logs don't show anything. CPU core temps are in the 60's so it doesn't seem like it would be heat related.

[jonault]

I looked at the logs from mid-January, this computer has been struggling with CPU work units for a long time, even before I switched it over to the v8 beta. At least with v7 it would try to resume from the last checkpoint and eventually it did finish the work unit; v8 seems to just give up at the first error so very few CPU work units get completed.

Frustrating that I had no idea there was a problem, but with 4 GPUs folding the CPU points are down in the noise. It took the GPUs acting up (and the reboots) to get my attention.