Plotting Performance

Chia Plotting
1

Hardware:
CPU: Ryzen 7 5800X (8-core/16-thread)
RAM: 32GB 3600 DDR4
NVME: Sabrent 2TB Rocket Q4 NVMe PCIe 4.0
Target HDD: 14TB USB Western Digital (x2)

I’m new to Chia Plotting/Farming, I built this PC in Dec 2020 and it seemed like it would be a good rig for some Chia plotting, so I thought I’d try it out. I’m running 5 plots in parallel via the following CLI commands using Powershell 7:

start-process powershell { cd ~\AppData\Local\Chia-Blockchain\app-1.1.6\resources\app.asar.unpacked\daemon\ ; Start-Sleep -s 0; .\chia plots create -k 32 -r 3 -n 25 -t C:\XCH\01 -d F:\XCH\Plots }

start-process powershell { cd ~\AppData\Local\Chia-Blockchain\app-1.1.6\resources\app.asar.unpacked\daemon\ ; Start-Sleep -s 3600; .\chia plots create -k 32 -r 3 -n 25 -t C:\XCH\02 -d F:\XCH\Plots }

start-process powershell { cd ~\AppData\Local\Chia-Blockchain\app-1.1.6\resources\app.asar.unpacked\daemon\ ; Start-Sleep -s 7200; .\chia plots create -k 32 -r 3 -n 25 -t C:\XCH\03 -d F:\XCH\Plots }

start-process powershell { cd ~\AppData\Local\Chia-Blockchain\app-1.1.6\resources\app.asar.unpacked\daemon\ ; Start-Sleep -s 10800; .\chia plots create -k 32 -r 3 -n 25 -t C:\XCH\04 -d F:\XCH\Plots }

start-process powershell { cd ~\AppData\Local\Chia-Blockchain\app-1.1.6\resources\app.asar.unpacked\daemon\ ; Start-Sleep -s 14400; .\chia plots create -k 32 -r 3 -n 25 -t C:\XCH\05 -d F:\XCH\Plots }

So 5 parallel plots with 1 hour staggered start up times. After running for about 2 days, I’m seeing an AVERAGE plot generation time of ~11 hours per plot. With 5 running in parallel, I’m getting about 10-11 plots per day. I had read that a setup with similar hardware could expect to generate plots in 4-5 hours, so I’m curious if my plot times are “normal” or if I need to tweak my settings, etc.

Any advise?!?

Thanks!

2

Casually observing performance, I’m noticing:

  • Sabrent NVME is consistently running at 350-400 MB/s sustained
  • CPU usage seems to be sporadic but averages 30-40%
  • RAM per plot is consistent around 2,000 MB (which seems low)

So is the sustained write speed of my NVME the limiting factor? Or should I be allocating more RAM per plot? Something else?

3

That’s 5x 269GB (1345GB) temp files on ur 2g (1.8g actual) C-drive? Even staggered isn’t that a bit risky for ur system drive? Also, if u added create commands up until (plot time-1) and once running, u get a plot every hr, isn’t that more like 24 plots/day?

1 Like
4

I’ve been watching free space on the 2TB C: drive and it hasn’t dropped below 800GB free so far, so it seems to be fine.

The original start times of the plots were 1 hour apart, but I’m still only running the original 5 plots in parallel (each is set to -n 25 to generating 25 plots in sequence) with plots taking ~11 hours each to complete while running in parallel, so 11 hours / 5 parallel plots = 2.2 hours per plot which is 10.9 plots per day.

If I run a single plot by itself, it completes in about 6.5 hours, which would generate a little less than 4 plots per day. Running parallel is definitely slowing down the individual plot time (11 hours vs 6.5 hours). So I’m trying to determine what the bottleneck is as well as the optimal settings to maximize plots/day I can generate on this hardware.

5

Obviously, the ssd is the bottleneck. Get another, run 7 plots on it, 5 on the other, for example, would help.

1 Like