@don256us
My suggestion with GPU core_22 continues to be one real cpu core for each gpu folding, 3 GHz or better. Best to have a core for the OS too. Also, suggest 4GB system ram per GPU minimum, 8GB preferred. My folding hosts are running Ubuntu on small 2.5" SATA SSDs, although a few are dual boot with W10 (for testing).
What a donor does with the remaining cores/threads is up to them. Just remember, most of the folding points come from the quick return bonus, and that ANYTHING that delays the return of the completed wu will have non-linear impact on credit received. Not sure if the clock starts when wu assigned or downloaded, but does run until upload completed and accepted.
The CPU is used to transfer data to/from the GPU, perform sanity-checks, checkpoints and download/upload data. Projects with frequent checkpoints require more CPU useage than slower ones. The GPU project "steps per checkpoint" is defined by the project owner. Fast GPU complete the steps more quickly than slow GPU, requiring more CPU too.
My folding farm is running on LGA 1155 motherboards (Asus P8Z77.V and P8Z77V-LK) with Ivy CPUs supporting PCIE Gen3, mostly e3-1265Lv2 (4C/8T), with a few e3-1220Lv2 (2C/4T) and i5-3470T (2C/4T). The cpus run around 3GHz and turbo to about 3.5 GHz. Typically have two GPUs running at x8x8 Gen3, although my 4070Ti are running in the primary slot at x16 Gen3 with a slow gpu is in the 3rd slot running at x4 Gen2.
On my hosts with quad core CPU (4C/8T), have Boinc Rosetta and World Community Grid running with resource setting 25% CPU (2 threads) and zero GPU.