I wonder if this deserves its own category. I, for one, am thoroughly tired of “critique my rig”
But man people love posting these things over and over…
TL;DR you want lots of cores, as fast as possible, for the lowest price. Then add enough fast drives, and enough RAM to cover the number of plots those cores can do. Done. GOOD LAWD, PEOPLE!
The absolute maximum number of parallel plots you can do on a given machine is equal to the number of cores and threads that machine has, added together, divided by two.
The absolute max is still a bit too high; on a 16c/32t machine, you’re probably not going to want to do 16 + 32 = 48, 48 / 2 = 24 parallel plots… but at least it offers a guideline for maximums…
How does adding cores and threads together make sense? Cores are physical, threads are virtual. Shouldn’t it just be at least 2 threads per plot and more if you want some speed gains?
Well, threads are like partial cores. Remember hyperthreading has been around a long time now and it’s pretty sophisticated at this point. So I’d say a thread is like half a core, as a rough rule of thumb, and I think that’s a fairly accurate statement.
Plotting is kind of complicated though. And people like to talk about their cars.
Core speed, thread coherency, CCX barriers, L1 & L2 cache sizes, RAM speed, multi-channel RAM, SLC vs TLC vs MLC vs QLC, on-die RAM & SLC cache cliffs, DC vs consumer drives, disk bandwidth, controller thrashing, PCIe lane contention, DMI bus vs direct-to-PCIe, bifurcation vs PLX.
But yeah, most people are stressing about “hai! how I mine 4 fish!” on a POS i7 and 16GB RAM.
No, I object to your point. You can try to use the maximum number of threads and start 2 or more tasks at the same time. Confirm the results. You will be surprised!