Often, after I've had the computer unplugged to work on its insides or after recovering from some sort of crash, when I go to start up GPUGrid WUs, my GPU will swing wildly from max clock speed to min clock speed resulting in very slow processing.

I've found that I can fix this by switching my GPUs over to another project such as Prime Grid temporarily and when I come back (after perhaps 20 minutes/2 WUs) the clocks will stabilize at max speed for GPUGrid.

I'm just curious if anyone else has ever seen this and what may be the cause of it. I'm running 2 GTX680s and usually it's just one of the cards that will do this.

Only seen this behavior before with bad WU's. What WU types were running?

Odd that it only happens after you have had the case open. Does it not also happen after a restart or suspend/resume crunching?

Have you tried using prefer maximum performance in the NVidia control panel (Windows) or X Server (Linux)?

---

FAQ's

HOW TO:
- Opt out of Beta Tests
- Ask for Help

The last time this happened (yesterday) I believe I was running SANTIs, but the type of WU doesn't seem to matter. I can usually predict with almost perfect accuracy that under the conditions I've described that I'll see this issue after resuming work regardless of the WU type (though I only run Long tasks).

I never seem to have a problem when I suspend crunching or close out of BOINC altogether and then start it up again. A normal shutdown or reboot won't have any effect either. I have Nvidia set up for maximum performance.

I don't know why switching to another project fixes it. Almost seems like the system has "forgotten" what it's supposed to do and running something else for a few minutes reminds it and fixes the problem for GPUGrid.