I would think it would have more to do with heat/loop saturation as opposed to cores not being available to the GPU's. The hotter the loop temp and conversely the GPU temps, the lower PPD you will put out. I checked my office this morning (Had the windows open) and the loop temp was 15c, the 2080Ti was @ 19c @90% load generating 6.4 million PPD.I noticed something I thought was unusual over the holiday weekend. I decided to run the cpu and gpu slots with the cpu set to 12 threads (out of 16). I got 7.1 mil (24hr) with 50 work units, I turned off the cpu slot and the next 24hr average was 8.5 mil with 31 work units. 4 threads should be enough to feed a 2080ti to it's potential, shouldn't it?
I didn't waste any more time messing with it since I am about to set up the linux machine, but it makes me wonder if I made a mistake thinking that a 6/12 processor would be enough to feed 4 cards (eventually). Maybe it will be as long as I don't do cpu folding.
You and I are in the same boat, currently looking for a setup to run my 4x2080 Ti's for folding. It should be fine as long as you do not run the CPU for folding. I would be more concerned about the PCI-e lanes, which is why I have a post looking for a used threadripper build. I assume the hit to a Linux based system will be less that windows and even a x4 lane doesn't have THAT stiff of PPD loss. Still if we are talking 20% lost ppd between x4 and x16, I'm going to try and get that TR build.My 2080ti is air cooled but the cpu is on an AIO (that needs dusting fo sho). My main concern is if a 5930k is going to be enough to run [email protected] on linux with a 1080ti, 2070, 1660, and eventually my 2080ti. I didn't think I would need much from the cpu if I provided 2 threads per gpu, but that may not hold true anymore.... and I'm not sure where I read that "rule of thumb".