GPU stuck at half clock speed while using CUDA in cycles.

Support Technical Support
1

I tried cycles for the first time and was curious to see how much faster my gtx 460 would be over my cpu (fx8350), and was slightly puzzled when I saw that it was a couple of seconds slower. So I opened up my monitoring program to see if my GPU was being fully utilized and found that the core clock speed is always idling at 405MHz instead of the 763MHz that it’s supposed to be under load. The gpu utilization was at 100% so clearly the full speed thing is not triggered by load, but rather some kind of flag that 3d applications have.

Does anyone know how I can force/ use a profile to make my card hit full clock speed while blender is in use?

2

I’m using Linux and a GT640, but I assume theres something similar in other OS’s.
Within the system settings theres an option supplied by nvida called “x-server settings” which allows either “adaptive” or “prefer maximum” under the GPU settings.
With this I can force the card to operate in a high mode all the time with not to much hit on the temp or fan speed.

3

Thanks, I figured it out before my post got approved. It was the first thing I tried, but didn’t work initially because the settings don’t take effect until you restart.