Cant really get the best out of the 3990x , Any advice?

Chia Plotting
1

I finally got to build a monster machine but not able to produce a lot of plots daily using the swar plot manager.

AMD 3990x , 256GB ram, 8x 2TB NVME ( Corsair MP600 )
MB gigabyte tx40 designare + AORUS Gen4 AIC Adaptor for 4 NVMe PCIe 4.0
Destination drives are all 7 internal SATA HDD and one external.

What would be the best config for the swar plot manager ?
so far this is what i got and i am barely making 50 plots a day.
the first batch of job, p1 takes about 2 hrs, but the following p1 could reach almost 7hrs

Global :
max_concurrent: 48
max_for_phase_1: 24
minimum_minutes_between_jobs: 1

  • name: Job 1 to 8 ( one job per drive )
    max_plots: 999
    farmer_public_key:
    pool_public_key:
    temporary_directory: D:\temp
    temporary2_directory:
    destination_directory: I:\plot
    size: 32
    bitfield: true
    threads: 4
    buckets: 128
    memory_buffer: 4000
    max_concurrent : 6
    max_concurrent_with_start_early: 6
    initial_delay_minutes: 0
    stagger_minutes: 60
    max_for_phase_1: 2
    concurrency_start_early_phase: 4
    concurrency_start_early_phase_delay: 0
2

I’ve got the MSI TRX40 also with a 3970x and 128GB. I’ve got 2x 2TB Samsung NVME on the mainboard and then have 4x 1TB Samsung and 4x 1TB Timetec (junk) in two Dell PCIe 3.0 (couldn’t find any 4.0) cards. Anyhow with that setup I do between 85-90/day. I haven’t spent much time tuning or tweaking this, so I believe I can get more out of it. The staggers and delays are pretty much just random from what I had read on here lol…

Here’s my SWAR Config:

max_concurrent: 50
max_for_phase_1: 13
minimum_minutes_between_jobs: 7

2x jobs for the Samsung 2TB’s:

  • name: 0-Samsung_2TB
    max_plots: 999
    temporary_directory: /ssd/ssd0
    temporary2_directory:
    destination_directory:
    • /dest/dest0
    • /dest/dest9
    • /dest/dest10
    • /dest/dest11
    • /dest/dest20
    • /dest/dest21
      size: 32
      bitfield: true
      threads: 5
      buckets: 128
      memory_buffer: 8000
      max_concurrent: 4
      max_concurrent_with_start_early: 6
      initial_delay_minutes: 0
      stagger_minutes: 35
      max_for_phase_1: 2
      concurrency_start_early_phase: 4
      concurrency_start_early_phase_delay: 0

4x jobs for the Samsung 1TB’s:

  • name: 2-Samsung_1TB
    max_plots: 999
    temporary_directory: /ssd/ssd2
    temporary2_directory:
    destination_directory:
    • /dest/dest0
    • /dest/dest9
    • /dest/dest10
    • /dest/dest11
    • /dest/dest20
    • /dest/dest21
      size: 32
      bitfield: true
      threads: 5
      buckets: 128
      memory_buffer: 8000
      max_concurrent: 3
      max_concurrent_with_start_early: 4
      initial_delay_minutes: 5
      stagger_minutes: 35
      max_for_phase_1: 1
      concurrency_start_early_phase: 4
      concurrency_start_early_phase_delay: 0

4x jobs for the Timetec 1TB’s:

  • name: 7-Timetec
    max_plots: 999
    temporary_directory: /ssd/ssd7
    temporary2_directory:
    destination_directory:
    • /dest/dest0
    • /dest/dest9
    • /dest/dest10
    • /dest/dest11
    • /dest/dest20
    • /dest/dest21
      size: 32
      bitfield: true
      threads: 5
      buckets: 128
      memory_buffer: 8000
      max_concurrent: 2
      max_concurrent_with_start_early: 3
      initial_delay_minutes: 10
      stagger_minutes: 60
      max_for_phase_1: 1
      concurrency_start_early_phase: 4
      concurrency_start_early_phase_delay: 0
3

thank you for your reply.
if i may ask, how long it takes you to do one plot. say on the first run and , the subsequent ones after ?

4

The first few plots after I clean restart will fly through in just over 4 hours. Once it’s rolling and has 30+ plots going it’s closer to 8-8.5 hours each. Here’s an example of a plot that just finished:

0-Samsung_2TB 32 ba60fbb 663343 2021-06-03 15:43:37 08:23:41 5 02:57 / 02:38 / 02:27 / 00:14