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

@madMAx43v3r Good morning Max. I believe I have Chia 1.8.0 Giga10 working correctly now, but I’m getting quite a few warnings like these below:

Are these anything to be concerned about? As always, thanks in advance.

Update: Well I thought it was fine as it appeared to be farming, but it hasn’t submitted any partials since updating to giga10 (almost 3 hours). I guess I’ll go back to Giga7 which I know works.

Update 2: Submitting partials again using Giga10. I went back to my old config.yaml. Not the new one that I created with Chia 1.8.0.

1 Like


But the error about G1 non-infinity element must start with ... usually happens when you mix incompatible versions…

Make sure to completely stop any previous chia services before starting Gigahorse, close the GUI and run chia stop all -d in the terminal.

And then when you start Gigahorse there should not be any messages about “already running”.

@madMAx43v3r I know that you said block validation time WARNINGS were nothing to worry about, but why am I constantly getting them? What block validation time would be of concern? I’m pretty sure that I never used to get them before upgrading to the latest version.

Block validation time has nothing to do with farming or compressed plots, you need to ask Chia team about this.

i see alright thank you very much

I have a 1650 Super with 4GB of VRAM, and 304 TiB of uncompressed plots, is this GPU suitable for farming C8?

Will it cope with K32, K33 and K34 size plots and for how long considering plot filter changes?

It will be in my headless home server running Windows 10.

Or am I just better off using a 3060 12GB (non Ti)?

Plotting will be done on a separate machine using a 3080.

Yes, it should be able to handle 500 TiB (physical size) of C8 even at plot filter 256. However K32 is the max, due to 4G VRAM.

1 Like

What is the point to make k33+ if k32 is the most profitable one since it supports higher compression level?

Bragging rights :wink:

But also higher K size has less disk IO, if that is an issue for you.

And of course the ever lasting fear of K33 being the minimum some day.


i see i see bragging rights we live for :upside_down_face:

let me repeat my another question above, i’m not sure you were answered me. plots check proof time doubles with next compression level, see screenshot above C7,C8,C9. is it all good, or C7 has better chance to win?

Yes of course, higher compression = more compute.

As long as you can compute fast enough it’s not a problem. C7+ is only intended for GPU.

well, is 2000ms is overkill for c9? it’s not fast enough, right?

Why would it not be fast enough? Anything below 20 sec for a full proof is fine. Quality lookups should be below 5 sec.

ok let me rephrase, would better hardware farming same plot size/compression more winnable because of less proof found time?

No, unless the slower hardware is too slow and produces stale partials.

Thank you for the reply, I’ll go with the 3060 then, I don’t game much these days so I’ll keep the 1650 in my main PC (3080 won’t fit)

By using a combination of plot sizes I could fill more space on the drives, that’s the main reason I used a combination of plot sizes.

1 Like

i’m not in the pool, so no partials

and yet most of them should be c9 to achive best farm size

C9 is usually less profitable due to higher compute cost