It would be interesting to see how that CPU demonstrates the difference between the “chia.exe” code and the “chia_plot.exe” code (the latter being madmax).
Of course, to take advantage of your threads, be sure to apply the “-r” option with a value that utilizes your ample supply of threads (perhaps “-r 250”, which leaves a few threads available for OS stuff).
Assuming you do not have a slow temp partition / bottleneck, madmax will probably be 20 or 30 times faster (just a wild guess). Perhaps even 50x faster?
If you want to get it to really plot efficiently, you will need to get much more RAM at the highest speed it supports (that is where the cost really is), and potentially a RAID0 NVMe for temp1. Otherwise, all those cores will be just waiting for something to do.
You may also check the CPU cooling part and upgrade that, as when you get it really running (choking all 64 cores), it may start thermal throttling.
The CPU in this thing is water-cooled from the factory. Never seen that before! I’ll be using a RAID0 array for temp1,
I have an add-in card that will set up four NVMe drives at x4x4x4x4 in a x16 slot. Plus some memory, I’m looking at about $1500 to fill up f out of the six DIMM slots. Might just get it plotting to disks and see how it goes, if it’s going to sleep waiting for I/O then I might see if I can a cheaper way around.
You had a great bargain for the price, but I see the board is limited on the max memory, it seems only supports 6 x 32gb modules, so it can’t plot only in ramdisk.
Try an optane ssd if you can get one.
Think some of the xeon skylake sp boards support more ram, if fewer faster cores work just as well. But may cost many times more than what you payed.
Sounds a lot for Ram, can’t you find used stick’s for a lot less?