[OPTIMIZING-PLOTTING] I need you expertise help

Chia Plotting
21

I have now tested different three different configs

  1. k34, high custom RAM and high threads setting
  2. k33 plot size, 4GB RAM and 4 threads
  3. default k32 plot and the default RAM and threads setting

My experience so far is that the default setting are LOT slower then making a custome config both option 1 and 2 have beaten the default k32 setting setup.

On option 3 the default setting I have now got only 19% after 9 hours of running time :open_mouth:

22

Now it looks like youā€™re comparing apples and pears. To do a consistent experiment, use the same amount of RAM and threads for all 3 sizes.

In several threads here people seem to have found 4 threads is optimal, so that could be why youā€™re seeing good speeds on your k32.

Also if youā€™re doing all of these at the same time, then they do influence your test results, if a k33 is deeply crunching away when your k32 is starting, also wanting to crunch hard, then the one that started first might just win the resource competition.

2 Likes
23

I will do more testing, but what I shared is as far I have come.
My ā€œmindsetā€ was that fixing the config setting with higher RAM/THREADS setting for the start would be the optimal way. " and found it strange that I finished my k34 plot faster then my k32 plot.

After my k32 plot is finished (option 3) I will run it like you saying a valid ā€œexperimentā€ should be.

k32 with 4GB RAM and 4 threads
k33 with 4GB RAM and 4 threads
k34 with 4GB RAM and 4 threads

To compare GB per HOUR on each of them.

When it comes to

so that could be why youā€™re seeing good speeds on your k32

It was the opposite I ment, I did not see good performance with the standard settings.

24
1 Like
25

Thanks for sharing .

1 Like
26

https://plot-plan.chia.foxypool.io/
Very nice tool.

1 Like
27

image
I saw in your video that you had 10000 seconds on phase 1. Do you have any clue why my phase 1 takes 5X your time?

28

Itā€™s not me on the video, but I can get phase 1 in ~13000 seconds.
At first glance, you are using less CPU than me. Here is my plot setup and timings:

Buffer size is: 3390MiB
Using 128 buckets
Using 2 threads of stripe size 65536

Starting phase 1/4: Forward Propagation into tmp files... Tue Apr 27 20:12:56 2021
Computing table 1
F1 complete, time: 89.087 seconds. CPU (159.53%) Tue Apr 27 20:14:25 2021`
.....
Forward propagation table time: 1794.312 seconds. CPU (150.740%) Wed Apr 28 00:00:08 2021
Time for phase 1 = 13632.088 seconds. CPU (142.060%) Wed Apr 28 00:00:08 2021
....
Time for phase 2 = 3566.634 seconds. CPU (98.580%) Wed Apr 28 00:59:34 2021
....
Time for phase 3 = 8122.251 seconds. CPU (95.850%) Wed Apr 28 03:14:57 2021
....
Time for phase 4 = 686.535 seconds. CPU (83.880%) Wed Apr 28 03:26:23 2021
Approximate working space used (without final file): 269.333 GiB
Final File size: 101.349 GiB
Total time = 26007.511 seconds. CPU (120.130%) Wed Apr 28 03:26:23 2021

Cpu specs:
Core i7-7700K@4.5Ghz, 16GB DDR4 2666, plotting drive is a NVMe SK Hynix 1TB.

1 Like
29

image
Its getting better.

30

I use 4 threads -r 4 and noticed a 12% speed increase on my MacBook Pro 2020 16"

1 Like
31

After a few days of trying different configs it looks like
image
is the best time I can finish PHASE 1 on.

On thing I notice, if I run parallel plotting my time for phase 1 goes up by 2x.
Do anyone have experience with this? that time goes up by using parallel option ?

32

Hi, How have your numbers been? I have the same CPU i7 7700 and are looking for folks that have similar systems to benchmark

Iā€™ve only been plotting for 3 days but to date Iā€™m using a 3 parallel with 135mins stagger in between and are getting 20-21k seconds per plot, havenā€™t gotten to the plot per day yet and Iā€™m still plotting in GUI manually

Someone in reddit said this

i7-7700k 4core, 32GB ram 3200mhz cl16, Firecuda sata drive - plot every 115 minutes. Here never more than 5 in parallel. 2 threads.

I wonder whether heā€™s talking about total plot time or delay here and the 5 parallel is 5 plots with 115 mins stagger each, so at the peak times there will be at most 5 plots running? with 135 mins stagger Iā€™m getting at most 3 plots running with only 1 (the last one) at the first phase.

@X_TECHNOLOGY_LIMITED sorry to hijack your thread

33

Is there any different between k33 and k32 except size aspect? what should we plot? thanks

34

Here is a table of my experiments with different settings. The top 4 are with the Samsung ā€œRapid Modeā€ turned on. (Intel Core i7 6700 (4T/8C), 16GB ram, temp Samsung 870 EVO 1TB SATA, dest Synology NAS)

Top 1 on 1.1.6, others on 1.1.5

Phase1Cpu Phase2Cpu Phase3Cpu Phase4Cpu Phase1Seconds Phase2Seconds Phase3Seconds Phase4Seconds CopyTimeSeconds TotalSeconds Threads Buffer Buckets
2,661 0,861 0,863 0,521 02h 24m 51s 01h 11m 02s 02h 37m 32s 16m 15s 15m 26s 06h 29m 41s 6 6780 MB 128
2,147 0,908 0,904 0,661 02h 52m 49s 01h 17m 15s 03h 02m 40s 19m 29s 15m 26s 07h 32m 14s 4 9000 MB 64
2,126 0,895 0,885 0,789 03h 13m 24s 01h 19m 23s 02h 55m 25s 14m 23s 15m 26s 07h 42m 37s 4 6780 MB 128
2,165 0,917 0,925 0,644 03h 21m 42s 01h 26m 59s 03h 17m 48s 18m 34s 15m 25s 08h 25m 05s 4 6780 MB 64
1,942 0,945 0,851 0,422 03h 29m 59s 01h 24m 23s 03h 20m 14s 27m 55s 15m 30s 08h 42m 33s 4 6780 MB 128
1,476 0,935 0,894 0,706 03h 47m 12s 01h 29m 35s 03h 17m 33s 22m 23s 16m 50s 08h 56m 44s 2 4608 MB 128
1,985 0,96 0,867 0,448 03h 29m 48s 01h 39m 17s 03h 33m 08s 28m 45s 15m 43s 09h 11m 00s 4 8192 MB 128
1,408 0,933 0,878 0,49 04h 15m 47s 01h 29m 01s 03h 34m 10s 26m 55s 15m 37s 09h 45m 55s 2 3390 MB 128
1 Like