Swar Job Config for 3*1TB SSD and 10900K. Is it Worth Striping?

mehdizare · June 4, 2021, 8:32am

Hi everybody. I have a system with specs below and recently started plotting Chia.
CPU: INTEL I9 10900k
M.B: ASUS Z590-F GAMING WIFI
RAM: GSKILL 64 GB BUS 4000 TRIDENT
SSD: 3* (M2 980PRO SAMSUNG 1TB)

I wanted to know if it’s worth it to stripe the three SSDs? I’m currently plotting about 21 plots a day with three job configurations below:

name: 1
max_plots: 9999
farmer_public_key:
pool_public_key:
temporary_directory: E:\temp
destination_directory: H:\Final
size: 32
bitfield: true
threads: 2
buckets: 128
memory_buffer: 5959
max_concurrent: 3
max_concurrent_with_start_early: 6
initial_delay_minutes: 0
stagger_minutes: 1
max_for_phase_1: 2
concurrency_start_early_phase: 5
concurrency_start_early_phase_delay: 0
temporary2_destination_sync: false
exclude_final_directory: false
skip_full_destinations: true
unix_process_priority: 10
windows_process_priority: 32
enable_cpu_affinity: false
cpu_affinity: [ 0, 1, 2, 3, 4, 5 ]

name: 2
max_plots: 999
farmer_public_key:
pool_public_key:
temporary_directory:
- F:\temp
  destination_directory:
- H:\Final
  size: 32
  bitfield: true
  threads: 2
  buckets: 128
  memory_buffer: 5959
  max_concurrent: 3
  max_concurrent_with_start_early: 5
  initial_delay_minutes: 0
  stagger_minutes: 180
  max_for_phase_1: 3
  concurrency_start_early_phase: 5
  concurrency_start_early_phase_delay: 0
  temporary2_destination_sync: false
  exclude_final_directory: false
  skip_full_destinations: true
  unix_process_priority: 10
  windows_process_priority: 32
  enable_cpu_affinity: false
  cpu_affinity: [ 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23 ]
name: 3
max_plots: 999
farmer_public_key:
pool_public_key:
temporary_directory:
- G:\temp
  destination_directory:
- H:\Final
  size: 32
  bitfield: true
  threads: 2
  buckets: 128
  memory_buffer: 5959
  max_concurrent: 3
  max_concurrent_with_start_early: 5
  initial_delay_minutes: 0
  stagger_minutes: 180
  max_for_phase_1: 2
  concurrency_start_early_phase: 5
  concurrency_start_early_phase_delay: 0
  temporary2_destination_sync: false
  exclude_final_directory: false
  skip_full_destinations: true
  unix_process_priority: 10
  windows_process_priority: 32
  enable_cpu_affinity: false
  cpu_affinity: [ 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23 ]

Yae · June 4, 2021, 10:52am

You never want to raid0 ssd’s for Chia, unless it is the only option to fit plots (e.g. 2x500GB can fit 3 plots when striped, only one each without.)

Voodoo · June 4, 2021, 11:57am

Why not? Many people are using various raid setup for their temp drives.

I think that will be worth it with 1 TB drives.

It gives you more space to work with and makes it easier to setup good staggering times on the plots.

I think that way you should run 8 plots concurrent, 4threads, max 4 in phase 1. That way you should pass 30+ a day easy

10900K should be able to finish plots quite fast when running like that

Yae · June 4, 2021, 12:02pm

Not all raid controllers (hardware and software) correctly pass the TRIM commands to the drives.

A Raid 0 will just make sure every operation is held up by the slowest in the array. For SSD drives the difference can be huge because of extremes in write amplification and garbage collection.

mehdizare · June 5, 2021, 2:17pm

Thank you for your answers. I tried the striping but not that much change happened. My biggest problem now is that the system freezes frequently. I have to completely turn off/on the system. Any similar experiences?
I decided to run the plotting with Chia app in three plots for each NVMe 2.0 and see if the freezing happens again.

helldorado · June 5, 2021, 5:49pm

Can you show global configuration thanks.

mehdizare · June 5, 2021, 7:22pm

max_concurrent: 11
max_for_phase_1: 8
minimum_minutes_between_jobs: 1

Voodoo · June 6, 2021, 10:31am

Freezing issues can very well be caused by driver issues.

Be sure to update bios and chipset drivers.
That last bit seems to have finally solved my bsod issues - knock on wood

From the configs i see, i think you are having too many plots in phase 1 at the same time. Really want to keep that as low as possible without having plots wait for eachother

mehdizare · June 6, 2021, 10:57am

I updated the BIOS with the built-in ASUS tool. Any solution for completely checking the drivers? By the way, I don’t see BSOD. The system just freezes and no mouse/keyboard/Ctrl+Alt+Delete works. I have to turn it off from power.
Anyway, I changed the jobs and lowered phase 1 as below. I started three jobs with the code bellow and no freezes after 8 hours yet:

name: 3
    max_plots: 999
    farmer_public_key:
    pool_public_key:
    temporary_directory:
    G:\temp
    destination_directory:
    H:\Final
    size: 32
    bitfield: true
    threads: 2
    buckets: 128
    memory_buffer: 3700
    max_concurrent: 3
    max_concurrent_with_start_early: 4
    initial_delay_minutes: 0
    stagger_minutes: 65
    max_for_phase_1: 2
    concurrency_start_early_phase: 4
    concurrency_start_early_phase_delay: 0
    temporary2_destination_sync: false
    exclude_final_directory: false
    skip_full_destinations: true
    unix_process_priority: 10
    windows_process_priority: 32
    enable_cpu_affinity: false
    cpu_affinity: [ 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23 ]

moosloos · June 6, 2021, 11:16am

just to show everyone how different things are in this space, i’ve been plotting 110-120 plots a day on nothing but raid0.

mehdizare · June 6, 2021, 11:45am

How many do you think I can do with these specs? I tried striping using windows built-in tools but nit that much of change happened.

moosloos · June 6, 2021, 12:41pm

26-28 easy. hyper optimized might get more.

one of my machines is a 10850 with 2x 970 EVOs and it does that.

i run ubuntu 20.04 LTS server, I use LVM to do my software raids. Others recommend mdadm as it has slightly higher throughput, but i don’t think thats the limiting factor for chia.

so my setup is all command line and i use Plotman, but SWAR is a fine manager

mehdizare · June 7, 2021, 6:22am

I ran the swar with the config above. The freezing problem is solved, but the system restarts after two hours of working occasionally. Any solutions? I updates the chipsets as well

GIGATOR · June 7, 2021, 6:52am

If event logs aren’t giving any hints it sounds like you need to go back to fundamental troubleshooting your system. Reseat all your connections. Run a full memory test. Run a 24 hour cpu stress test. consider if you might have a bad power supply and or motherboard. There are too many factors that can cause random system instability that you just need to start checking off.

As for raid 0 temp SSDs everything I’ve seen says yes, it’s all upsides with no down. It isn’t 2005, motherboards know how to handle SSD’s properly now.

mehdizare · June 10, 2021, 6:50am

Do you suggest any particular tool for checking system problems? I still randomly have freezes or restarts.

Voodoo · June 10, 2021, 9:15am

yes lot’s of things can be the problem. Also, depending on your local power grid, a UPS or AVR might also help.

I was having issues with BSOD/random restarts and since I update the X570 chipset drivers it seems to have gone away.

mehdizare · June 10, 2021, 9:43am

I’ve updated everything, even the firmware of my SSD drives. Do you think internet connection issues may result in freezes for swar and as a result, the whole system?

Voodoo · June 10, 2021, 9:48am

Swar can run fine offline, no internet needed.

I would look at non chia related issues, like suggested, run memtest64 or some other memory test, cpu stresstest, with no chia programs running and see what you get.

If it still fails then it really might be a hardware issue, such as bad Mainboard, PSU or …troubleshooting sucks. Have you tried the windows crashreports?

( Or Install Ubuntu )

mehdizare · June 10, 2021, 9:53am

You know, I started some memory and disk checks. but can’t find reliable software for CPU tests. and is there any useful tool to test the health of the motherboard?

FATSAC · June 13, 2021, 11:02pm

This system should easily be able to do 10 in parallel with the default chia plotter…

So assuming your final drive is an HDD and your temp is 3x1tb(raid0). Try this… it should start a plot every 30 min… which should be enough time for the plot to write to the HDD without being caught by the next plot… unless your HDD is only writing at 40MB/s or something… this would assume you are able to write at 100-200MB/s to the HDD.

name: 10_Plots+
max_plots: 9999
farmer_public_key:
pool_public_key:
temporary_directory: E:\temp
destination_directory: H:\Final
size: 32
bitfield: true
threads: 2
buckets: 128
memory_buffer: 4000
max_concurrent: 10
max_concurrent_with_start_early: 10
initial_delay_minutes: 0
stagger_minutes: 30
max_for_phase_1: 10
concurrency_start_early_phase: 4
concurrency_start_early_phase_delay: 0
temporary2_destination_sync: false
exclude_final_directory: false
skip_full_destinations: true
unix_process_priority: 10
windows_process_priority: 32
enable_cpu_affinity: false
cpu_affinity: [ 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18 ]