Hi, I've updated Fedora 8 x86_64 NVidia drivers to 180.22. The new workunits I've downloaded started up fine for several minutes. After about 30 minutes of crunching, the temperature for the GPU went down to idle, and the CPU usage went to 100%. It has been crunching like this since. Progress has been stuck at 0.8% too.

Why did CUDA stop using the GPU to crunch and switch to the CPU? What else is going on here?

I restarted BOINC and that seems to have helped. I'll check on it periodically while I'm doing homework tonight. I thought it was pretty weird too.

I have the exact same problem with Seti@home CUDA client, as well as with GPU Grid. The problem is either related to the BOINC manager, as I've noticed issues with the GPU client when CPU tasks switch, or it's a problem with CUDA that may be fixed in a future driver upgrade.

I am using Fedora 10 and was experiencing weirdness/problems, mostly compute error results. I finally used nvclock_gtk to crank the fans up to 100% duty cycle. Haven't had a problem sense.

Curt