Gigahorse GPU Plotter, GPU Farmer, Plot Sink: Boost Your Chia Farm by 47.2%! Plot in <2 Minutes

I had to run c7, with only 3 G of vram c8 plots crashed the farmer. I am also running a 1030 with 2G of VRAM that can only handle c6 My maths is fine, have another beer :beers::beers:

first of all, my concerns about C8 level is not true.
It seems buying 600 pcs of 8TB refurbished – very good prices – was not so good ideea, but i will make it work.
Thanks Max. You’re the best!!!

1 Like

that is wow! gonna do that too. i wish i could have a beer, its morning!

Yes, it seems that’s not the issue. As my 16 TB 7200 RPM are listed at 4W idle too.


I get an error on startup.
I restart - it works.
After a reboot, the error is back.

The problem is that the wallet freezes after a while and farming stops. After working for about 3-… hours, the wallet just starts to close.

C:\gigahorse>chia.exe start farmer
Daemon not started yet
Starting daemon
Traceback (most recent call last):
File “C:\GIGAHO~1\”, line 4, in
File “”, line 1027, in _find_and_load
File “”, line 1006, in _find_and_load_unlocked
File “”, line 688, in _load_unlocked
File “C:\GIGAHO~1\”, line 135, in
File “C:\GIGAHO~1\”, line 131, in main
File “C:\GIGAHO~1\click\”, line 1130, in call
File “C:\GIGAHO~1\click\”, line 1055, in main
File “C:\GIGAHO~1\click\”, line 1657, in invoke
File “C:\GIGAHO~1\click\”, line 1404, in invoke
File “C:\GIGAHO~1\click\”, line 760, in invoke
File “C:\GIGAHO~1\click\”, line 26, in new_func
File “C:\GIGAHO~1\chia\cmds\”, line 23, in start_cmd
File “C:\GIGAHO~1\asyncio\”, line 44, in run
File “C:\GIGAHO~1\asyncio\”, line 649, in run_until_complete
File “C:\GIGAHO~1\chia\cmds\”, line 64, in async_start
File “C:\GIGAHO~1\chia\cmds\”, line 42, in create_start_daemon_connection
File “”, line 322, in decode
UnicodeDecodeError: ‘utf-8’ codec can’t decode byte 0xc0 in position 29: invalid start byte

C:\gigahorse>chia.exe start farmer
chia_harvester: started
chia_farmer: started
chia_full_node: started
chia_wallet: started


OS Server 2022

told ya, it has something to do with python

What compression do u use for ur 16TB drives?

The size of HDD doesnt matter, C level is all about your compute. I use K31 C8 for MMX and K32 C7 for Chia, because I want to farm it all with my K2200.

On my friends ~18 PB farm we use K32 C8 with 10x A4000 GPUs.


How much system memory is in the box with the K2200 and what type of CPU?
That you farm both on the same box.

Well 192G RAM, but you dont need that of course. CPU is single E5-2470V2, but again also overkill. I just have 8x 16 TB HDDs.

This is my farmer test machine basically, running 24/7 with latest versions of MMX / Gigahorse / Flexfarmer.

Please tell me you are testing on Linux? Or is there a test one for Windows too, and which version are you testing?
I have on Server 2022 as I wrote a little higher errors and wallet crashes

It’s Linux. I do have a Windows Server 2022 as well, but I dont run it 24/7 since the license cost is crazy high (it’s a cloud VM). The stupid license cost more than the entire instance.

The license can be bought not at all expensive if not officially. About $30 for 16 Data Center cores - you just need to look.
Many windows and if it crashes I need to either stop using it or look for a reason, I want to find the optimal OS and build on it for 24/7/365

linux - ubuntu 22.04

If you don’t mind sharing that info.

What is the layout of that farm? Is that 10x harvesters each having a single A4000, or rather a single decompression box with all those GPUs serving all those remote harvesters?

Also, are those harvesters a single top loaded boxes with ~96 HD slots? If that is the case, what kind of temps are on those HDs?

Farmers are R730 with two A4000 each, and 4x 60 bay JBODs per farmer. 5 of that in total.

Temps are pretty bad right now, due to plotting heat output, over 40 C for the HDDs.


Thank you. Nice setup.

I am really looking forward for that external decompression module, as that is also a new way to move heat generating parts away from HDs (in addition to all the other benefits). I hope that it will come alive soon.

Is there a list/sheet floating around somewhere with farm sizes for different GPU’s ?

That memory usage seems to be at odds with the info here:
where it says C7 memory would be 1,663 MiB
Or are you farming K33?

Hi, what does “ans_encode_sym(): index out of bounds: 27844” mean? Windows 10.

C:\CUDA>cuda_plot_k32 -n -1 -C 8 -t r:/ -d \Desktop-110hbkm\j\ -d \Desktop-110hbkm\l\ -d \Desktop-110hbkm\v-s04\ -d \Desktop-110hbkm\w\ -A 3 -M 96 -S 3 -c xch… -f …
Chia k32 next-gen CUDA plotter - 3e00fa3
Plot Format: mmx-v2.4
Network Port: 8444 [chia]
No. GPUs: 1
No. Streams: 3
Final Destination: \Desktop-110hbkm\j
Final Destination: \Desktop-110hbkm\l
Final Destination: \Desktop-110hbkm\v-s04
Final Destination: \Desktop-110hbkm\w
Shared Memory limit: 89.3 GiB
Number of Plots: infinite
Initialization took 0.314 sec
Crafting plot 1 out of -1 (2023/04/01 14:00:48)
Process ID: 7392
Pool Puzzle Hash: 12bfde1

Crafting plot 15 out of -1 (2023/04/01 15:25:41)
Process ID: 7392
Pool Puzzle Hash: 12bfde1b32d93f11b1bb10a30ac44f0a30061efca306762c548899cfb9c3276f
Farmer Public Key: 8f4690f932d11f07fd185ea30102586e8b181d13429154f646c9113d2aaf8852dba5eb8309903d4ae0c078f60331c20a
Working Directory: r:/
Working Directory 2: @RAM
Compression Level: C8 (xbits = 8, final table = 4)
Plot Name: plot-k32-c8-2023-04-01-15-25-68d41aa2f7912bbaa6e07ebb42177f6d943bc8a005ef415ea4993db8a42e7ae6
[P1] Setup took 1.14 sec
[P1] Table 1 took 4.362 sec, 4294967296 entries, 16788092 max, 66594 tmp, 0 GB/s up, 7.79465 GB/s down
[P1] Table 2 took 26.577 sec, 4294678649 entries, 16790925 max, 66614 tmp, 1.20405 GB/s up, 1.91896 GB/s down
Flushing to disk took 44.653 sec
Started copy to \Desktop-110hbkm\l\plot-k32-c8-2023-04-01-15-19-745075e7edc9b6d79cf5060d9432302c708a8f3f899ce0f29bdd19648b978067.plot
[P1] Table 3 took 29.311 sec, 4294241306 entries, 16784293 max, 66742 tmp, 1.6375 GB/s up, 2.89994 GB/s down
[P1] Table 4 took 33.841 sec, 4293537474 entries, 16784223 max, 66557 tmp, 2.3636 GB/s up, 3.51645 GB/s down
[P1] Table 5 took 34.688 sec, 4292072601 entries, 16780175 max, 66528 tmp, 2.30551 GB/s up, 2.94051 GB/s down
[P1] Table 6 took 33.574 sec, 4289176188 entries, 16769358 max, 66658 tmp, 1.90495 GB/s up, 2.53173 GB/s down
[P1] Table 7 took 27.711 sec, 4283168091 entries, 16742392 max, 66526 tmp, 1.72983 GB/s up, 1.68706 GB/s down
Phase 1 took 191.632 sec
[P2] Setup took 0.328 sec
[P2] Table 7 took 7.399 sec, 4.31303 GB/s up, 0.0718002 GB/s down
[P2] Table 6 took 6.451 sec, 4.95378 GB/s up, 0.0823516 GB/s down
[P2] Table 5 took 7.845 sec, 4.07628 GB/s up, 0.0677183 GB/s down
Phase 2 took 22.182 sec
[P3] Setup took 0.818 sec
[P3] Table 4 LPSK took 7.035 sec, 3463945367 entries, 15576626 max, 61548 tmp, 4.62269 GB/s up, 7.2495 GB/s down
Renamed final plot to \Desktop-110hbkm\w\plot-k32-c8-2023-04-01-15-07-6c530a553066cfff1361ae44ed1bc939b3af17779efdd9b4c6a8ee6bc6adee3e.plot
ans_encode_sym(): index out of bounds: 27844
[P3] Table 4 NSK took 16.08 sec, 3463945367 entries, 13546431 max, 61548 tmp, 2.4075 GB/s up, 3.70244 GB/s down
[P3] Table 5 PDSK took 6.811 sec, 3529145901 entries, 13807299 max, 54767 tmp, 4.77311 GB/s up, 6.86393 GB/s down


and this: WritePark(): ans_length (859) > max_ans_length (858) (y = 4, i = 621)

[P3] Table 5 LPSK took 17.513 sec, 3529629154 entries, 14249363 max, 57089 tmp, 3.53851 GB/s up, 2.91214 GB/s down
[P3] Table 5 NSK took 13.782 sec, 3529629154 entries, 13802630 max, 56670 tmp, 2.86219 GB/s up, 4.31978 GB/s down
[P3] Table 6 PDSK took 6.251 sec, 3707853619 entries, 14510998 max, 57600 tmp, 5.19809 GB/s up, 7.47884 GB/s down
[P3] Table 6 LPSK took 12.631 sec, 3707853619 entries, 15090718 max, 60382 tmp, 5.08931 GB/s up, 4.0377 GB/s down
WritePark(): ans_length (859) > max_ans_length (858) (y = 4, i = 621)
[P3] Table 6 NSK took 14.436 sec, 3707853619 entries, 14497266 max, 60017 tmp, 2.8705 GB/s up, 4.12408 GB/s down
[P3] Table 7 PDSK took 7.149 sec, 4284698983 entries, 16763528 max, 66448 tmp, 6.13999 GB/s up, 6.53941 GB/s down
[P3] Table 7 LPSK took 14.085 sec, 4284698983 entries, 17203887 max, 69090 tmp, 5.07778 GB/s up, 3.62089 GB/s down