Doing around 50 Plots/d. But P1 is around 4hrs and seems to be the bottle neck. Advice for improving stagger times and settings to maximize plots/d is very much appreciated! Thanks
Hardware: CPU: i9-10980XE 18-Core overclocked 4.2GHz all cores
RAM: 64 GB DDR4-2933
3xSamsung 970 EVO Plus 2TB
1xSamsung 970 EVO Plus 1TB
1xCorsair MP600 1TB
Running 5 Jobs (1 per SSD) now with global: max concurrent:18; max_for_phase_1: 10; 0 min between jobs
Jobsettings: mostly Standard
Job 1,2,3 - 2TB Drives: 3 Threads per Job, max for P1: 3, Stagger minutes: 10
Job 4,5 - 1TB Drives: 3 Threads per Job, max for P1: 2, Stagger minutes: 10
Here’s some thoughts. No particular order, and YMMV:
Only allow one plot in P1 per drive
Don’t overload the drives with I/O. In my experience, I get more plots per day without filling my drives to capacity with running plots. 3 or 4 on my 2TB Corsair MP510 is about the sweet spot – any more and it grinds to a halt. Plotting 3-4 (staggered) in parallel takes over twice as long per plot as running just one by itself, but I’m getting ~3.5 instead of 2.x in the same amount of time.
You can lump your drives in most plot managers, for instance Swar and plotman both allow lists as Temp Drive inputs. They will fire plots on drives in the order down the list as they are called (from your stagger or phase limits). This will help you get the most from combining the coexistence of both the “stagger time” and “phase limit” parameters. It didn’t click for me until I tried it, but it’s quite useful.
Stagger your plots more than 10 minutes. It’s more efficient to have plots spread out equally than to have them performing near-identical tasks simultaneously. Ideally you want each drive to finish a plot and be right in sync for starting the next plot while maintaining the maximum number of plots each drive can support. This also ties in to the revelation I had regarding stagger vs. phase limits – phase limit is for making sure the drive has completed the phase before firing another plot on THAT drive. Stagger is the wait time after firing a plot on Drive A before starting one on Drive B. Very useful for avoiding final-copy-to-spinning-rust bottlenecks.