Does it finish under 1 hour? / Not exactly for the average Joe!
A simple test done on an i7-8700K clocked at 4.3 GHz, Evo Plus (Gen 3), the “phase 1” finished in 55 minutes (which is impressive comparing to the current version of chia plotter).
What seems to work the best so far:
- Bucket size: 512 (i.e. -u 512)
- Temp1 disk: a fast NVMe (25% of all writes)
- Temp2 disk: a RAMDisk or your BestDisk (75% of all writes)
What does choosing 2 temp drives do?
Files are split over both drives, the bigger .TMP files in the 1st drive and the rest smaller pieces (buckets) into the 2nd drive. During the phase 1, 2nd drive takes the .75% workload while the 1st drive takes .25%. After the phase 1, the workshift goes to 1st drive between 15 and 25%, the 2nd drive stays idle.
What about RAM?
RAM allocation does not matter. It takes “only” the default RAM needed per plot. Just above 3 GB of RAM. You can set it to 10 GBs or 20 GBs in the settings but it does not change anything.
And about CPU threads?
In phase 1, it takes “all” the CPU power your system got, all at 100%. Say, you have 12 cores 24 threads, even you set it to use 16 threads, it takes every thread usable. After the phase 1, now the CPU usage bounces up and down, but it’s way better than the official chia plotter, still multi threaded although the constant 100% CPU is ended after the 1st phase.