Conclusion
7% uplift at 3200mhz versus 2666mhz on the same RAM rated for 3200mhz. 2666mhz gave me 9h 30m plot times whereas 3200mhz gave me 8h 50m plot times. See settings below for full details.
Stumbled upon this because RAM was reset to 2666mhz after upgrading BIOS and I wondered why my plot times increased.
Settings
15 plots split across 5 queues with 10 minute delay in each Queue to allow for final xfer to HDD
Final destinations are 5 different HDD
2 threads
6750MB Memory
I have been following that thread. I figure RAM optimization goes beyond 5950x so I wanted to give it its own thread.
I am still building a simple Python script to automate. Manually, I set off 15 plots twice a day, so 30 plots/day. Once I automate, I could hit ~40 plots/day. Once I upgrade my NMVE to 970 pro 1TB and get my NVME to PCIE card so I can run 4 instead of 3, I think that number will go up. Hoping for 50, but no idea.
I am not certain what my bottleneck is right now, but the Firecuda 520 2TB burst at 4500mbs but then fall to 500mbs for sustained writes, so I am guessing those are the bottleneck.
I don’t think trying for 4 cores to burst through Phase 1 and trying wonky staggering is the way to go with 5950x. Once I get more connectors for my HBA card to SATA my ideal settings will be the same as now except 15 plots going to 15 different HDD destinations.
I agree. I’m tempted to try running the RAM faster than the current 3600MHz, but I don’t want to break plotting.
Sounds about right, you need to throw more IOPS into the mix, because even at 52 (was 51) plots a day my 5950X barely reaches 70% utilization so the CPU can do more if the SSD setup is capable enough.