CPU thread allocation and iGPU question

[CaptainHalon]

Using a 10850K here, which is 10C/20T. When I created my CPU slot, I set the threads to 10. My CPU usage is roughly at 77%, although some of that is for my Nvidia GPU. Looking at Windows task manager, FahCore_a8.exe is using around 67%. Does that sound right for 10 threads out of 20? I'm fine with the usage, just wanting to understand it better.

Also, on a separate note, my iGPU is currently disabled. Would it be worthwhile to enable it and fold on it? No idea what to expect with that.

Thanks.

[PaulTV]

About the iGPU, it's possible to enable it (another recent threat, also talks about enabling Intel iGPU: viewtopic.php?f=61&t=37150), but there currently aren't any jobs available for it.

[PaulTV]

About multi-threading... A multi-threaded core with two threads is somewhat faster than the same core with a single-thread. How much faster really depends on the application, but on average it's about 20 to 30%. In your case, you should at least keep one thread per active GPU, and one for the OS itself. Note that increasing the number of threads only applies to the next job, decreasing the number is immediate.

[CaptainHalon]

Good to know about the immediate change on decrease, but delay on increase. Never saw that in the config guide. I'm currently playing with the thread count to find the sweet spot for my usage, which I think will be somewhere between 12-16.

Doesn't sound like it'll be worthwhile to enable the iGPU.

[rkv_2401]

Make sure you also try it with CPU folding off if you're on Windows. Me and some of my team members all discovered that GPU PPD production was highest with CPU folding off (on a bunch of CPUs ranging from 6-core i7 10750Hs and R5 3600s to a 16-core 3950x, tried all combinations from just 4 threads CPU folding to [1 - total no. of threads]) with nVidia GPUs. The one thread per active GPU thing still works well with Linux.