How's the performance of Core Xeon?

Chia Plotting
42

NP. It appears it’s gone up in price, and is currently on back order at Newegg. No surprise to either right now, but this is the exact one I have:

And Plotman analysis from it for yesterday:
±------±—±------------±-------------±------------±------------±-------------±--------------+
| Slice | n | %usort | phase 1 | phase 2 | phase 3 | phase 4 | total time |
+=======+====+=============+==============+=============+=============+==============+===============+
| x | 30 | μ=100.0 σ=0 | μ=8.7K σ=142 | μ=4.7K σ=25 | μ=9.1K σ=26 | μ=600.1 σ=20 | μ=23.0K σ=158 |
±------±—±------------±-------------±------------±------------±-------------±--------------+

1 Like
43

I bought my machine based on your video, but unfortunately, I am not getting anywhere the same results as you. I am wondering if you can give me some tips?

I got a Dell T7810 with 1x E5-2680v3 (12c/24t), 32GB RAM DDR, and running 2x 1TB Corsair MP600 (one connected to a 16x PCIe slot and the other to an 8x slot). Not running them on raid0, and running this on Windows.

Only getting 8 slots/day. It seems to me that my issue is my 2TB worth of NVME as I haven’t even been able to use all my memory and the CPU is around ~80% during the day.

My Swar config is as follows:

global:
  max_concurrent: 9
  max_for_phase_1: 4
  minimum_minutes_between_jobs: 3

  - name: mp600_1
    max_plots: 999
    farmer_public_key:
    pool_public_key:
    temporary_directory: F:\
    temporary2_directory:
    destination_directory: Z:\plots
    size: 32
    bitfield: true
    threads: 3
    buckets: 128
    memory_buffer: 4000
    max_concurrent: 4
    max_concurrent_with_start_early: 4
    initial_delay_minutes: 0
    stagger_minutes: 60
    max_for_phase_1: 3
    concurrency_start_early_phase: 4
    concurrency_start_early_phase_delay: 30
    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: mp600_2
    max_plots: 999
    farmer_public_key:
    pool_public_key:
    temporary_directory: G:\
    temporary2_directory:
    destination_directory: Z:\plots
    size: 32
    bitfield: true
    threads: 3
    buckets: 128
    memory_buffer: 4000
    max_concurrent: 4
    max_concurrent_with_start_early: 4
    initial_delay_minutes: 0
    stagger_minutes: 60
    max_for_phase_1: 3
    concurrency_start_early_phase: 4
    concurrency_start_early_phase_delay: 30
    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 ]

Is there any tip you can give me to improve my setup? Should I buy 2 more 1Tb Corsair sticks?

Also, another question as I don’t remember you mentioning this in your video: when you complete the plot, do you copy it to the SAME nvme as destination (this would be an instant copy) and then run a script to move to your cold storage/NAS/etc? Or are you having your destination straight to the NAS?

Thanks for your cool videos :+1:

44

Plotting Phase 1 (And I believe Phase 3) prefer higher frequency CPUs - the more cores don’t necessarily mean linearly increasing production.

I think the older gen Xeons were in the ~3ghz range right? A 5Ghz chip would give you a lot more throughput.

45

Are you trimming your ssd’s? I trim mine hourly. Here’s my crontab:

22 * * * * bash -c "/root/myTrim.sh ssd.nvme.2G"
33 * * * * bash -c "/root/myTrim.sh ssd.sata.1G"
44 * * * * bash -c "/root/myTrim.sh ssd.sata.2G"
55 * * * * bash -c "/root/myTrim.sh ssd.sata.500"

And here’s my “myTrim.sh” shell script:

#/bin/bash

export T="/var/tmp/fstrim.$1.out"
export D="/media/$1"

date >> $T
/usr/bin/time -a -o $T /usr/sbin/fstrim -v $D &>> $T 
echo >> $T

That shell script assumes things are mounted like this:

/dev/nvme0n1    1.8T  1.3T  539G  71% /media/ssd.nvme.2G
/dev/sdh        1.8T  1.4T  409G  78% /media/ssd.sata.2G
/dev/sdf        469G  330G  139G  71% /media/ssd.sata.500
/dev/sde        916G  622G  295G  68% /media/ssd.sata.1G

And it creates log files here so that I can make sure “fstrim” is behaving correctly (and not too slowly):

-rw-r--r-- 1 root root 20K May 25 23:38 /var/tmp/fstrim.ssd.nvme.2G.out
-rw-r--r-- 1 root root 20K May 25 23:33 /var/tmp/fstrim.ssd.sata.1G.out
-rw-r--r-- 1 root root 20K May 25 23:47 /var/tmp/fstrim.ssd.sata.2G.out
-rw-r--r-- 1 root root 20K May 25 23:55 /var/tmp/fstrim.ssd.sata.500.out
1 Like
46

that’s correct, but that is not what my old Xeon has so a higher frequency won’t help my situation.
@SlothtechTV posted a build where he used a Xeon which I think is around 3Ghz and he’s getting 40 plots per day. I was hoping I could at least get 15 with my setup!

47

interesting, never thought about it!
I am on Windows, so let me find an equivalent trim script for powershell and give that a go! Thanks for this

1 Like
48

Optimize-Volume -DriveLetter H -ReTrim -Verbose

awesome!

Edit:
These are my jobs currently running. The nvmes are very busy (F:\ and G:\ - always around ~90%)
but as you can see the CPU and RAM are not super busy.
If I buy another 2x 1Tb sticks and place them on my PCIe slots, ill have a total of 4TB and will be able to run more in parallel, and then my problem will be memory.

image
image1685×683 141 KB