Gigahorse farming experience share

Have been plotting and using Gigahorse farmer and GPU plotter in the past two weeks. Previously using Bladebit cuda compression for around one month. Share some experience that might help other people.

  1. Bladebit Cuda compressed plots in any level will not generate any coins. I had around 8PB plots done using bladebit cuda compression C5, C6 and very few C7, running several harvesters and one farmer without pool (not working with any pool anyway, CPU would be 100% usage on any pool, server would freeze). Before compressed plots, I would get at least one block everyday, sometimes three or four blocks. With blade cuda compressed plots, running almost three weeks, no single block found. No errors and no warnings in the log.
  2. Then switched to Gigahorse farmer and plotter. 100Gbps network with many storage servers with raid0, a few GPU servers and each run multiple 3090 GPUs. Each GPU in virtual environment can do 3 minutes per plot by using NVME drive as temp drive. Using enterprise SSD drives can do 6 minutes, but the transfer to storage servers can take up to 30-180 minutes each plot once accumulating a lot plots, and plotter stops. On the other hand NVME drive can transfer each plot to storage servers in less than 3 minutes, no plot accumulation on the temp drive. My current utilization of GPU plotting capacity is close to 1PB per day. We have many 3090, full utilization would be around 4-5PB per day.
  3. After using Gigahorse farmer, hit one block less than 8 hours, obvious Gigahorse farmer is quite efficient. My harvesters (6PB none compressed and 3PB compressed) are all using CPU (2x E5-2695V2) to do farming, in the test period from June 11 to June 13 three days, hit total 4 blocks, and Gigahorse farmer took two 0.25 as fee, 0.5/8, the fee percentage is around 6.25%.
  4. Still have the deleted 6PB bladebit compressed space to fill, will update once all plots are done.

Well then you did something wrong…
It works with any pool, partials as normal.
Block wins confirmed by many. Just won one few days ago (etw 14 days).
Im running 5k c4 plots on flexpool atm. Cpu (i3-6100) usage is minimal, around 5-7W extra.


I think that means there’s something wrong with your setup. Would be curious to know what.


The bladebit compressed plots were done by Windows plotter under Windows Server 2022. I went through some posts, they say only Windows 11 working. That might be the reason. However, the bladebit developers seem lacking motivation to make it fully working soon. Just wasting time and a lot of resource, not going back yet, maybe after the final release.


Since replotted all my farm using bladebit cuda plotter, with c5 level, I was inmediately winning blocks, in fact more than before (almost 2 per day during the first 3 days).
So something was clearly wrong in your config.

1 Like

What is your plotter setting environment?

Ubuntu. Plotter, full node and harvesters.

I had the same issue. VRAM utilization of the harvester process was extremely low and it looked like it wasn’t even doing anything. I didn’t spend a ton of time looking into it but my best guess was a CUDA or kernel version conflict or something like that.

1 Like

ah IC, yeah server versions are base on win10 I think.

And yes…it’s taking quite long :sweat_smile: 4 months after announcement and not even in beta stage yet

1 Like

I have to say, Max did a good job with GH. I only wish he gave more of his fee to Flexpool.

1 Like