Folding Forum

GPUs do not have load balancing like CPUs have. GPUs also do not have throttling, to every GPU application will try to use all of the resources it needs without regard to any other application.

Folding on a GPU and rendering a game or even a Windows screen at the same time is like trying to watch 10 movies or 10 youtube videos at the same time. They all fight for limited resources. That's just the nature of the hardware for GPUs.

Thanks, 7im.
I guess some projects use a bit more than others. Other than a bit of video sluggishness, the WU completes just fine without errors or warnings.

I notice that when the video (GTX 970) start slow down, appears strong lag on the screen. When core 0x21 operate normally lag is negligible (tolerable). I work on Solid Works (CAD) at the same time.

My card is also powering my monitor. And i noticed that every core_21 slows it down it. I had 3 of 21 since my first post here, and all of them had this. While none of _18 were.

Again, this is a known behavior of GPU folding. It's just that higer end GPUs hide the lag issue better on the existing Cores. Only when a new core and larger work units came out did this become noticeable again on most every GPU.

And the recommendation has always been to run your monitor on either a low end card (like the one you just upgraded from) or the video built into the CPU processor.

I am now folding beta project 11401, and seeing sluggish video response for the first time with that project. It is actually quite moderate to severe. Could the issue be the 0.0.12 version of core 21 or a problem with a windows update? This card (GTX 960) has no issues running the project so far. It should complete in about 7 hours.

EDIT: I found that the WU was locked-up. I paused and restarted. It is progressing fine now and video response is back to normal. I will post in the beta thread when it completes.

When the UI becomes sluggish, it is very often the sign of a dying WU ... the one I call the "slow down bug", because the WU progress will slow down, until a bad state is triggered at next sanity check.

This bug should be fixed in future v0.0.13 of core 21, although no ETA given yet.