Gigahorse Compressed GPU Plot Client Released! Start Plotting Now!

Great, thanks for the confirmation.
Will start testing and replotting next week.

Thanks for all the hard work and availability to reply to the various questions from the community. Hope you’ll make enough via the fees to allow you to continue developing your product(s) in the future.

2 Likes

Once, Chia releases official support, we may do a study of rentability of the new optional format.

Regular plot farming is confirmed to be broken, posting an update soon.

1 Like

OK thanks! Also, I can’t seem to get the farmer to recognize the compressed plots, I have 4 plots on a drive, 3 compressed and 1 uncompressed, farm summary only shows the 1 uncompressed plot. The log shows that it was unable to open the file due to it being an invalid format. I double checked to make sure I am running gigahorse node, farmer, harvester. EDIT: Never mind, I was using chia start farmer instead of chia.bin, it looks like it is working on the new version, thank you!

This version works now to farm regular uncompressed plots as well:

That’s why I changed the readme to use ./chia.bin without activate.sh

You can still use . ./activate.sh and then chia ..., just have to remember not to forget it.

3 Likes

The cuda plotter was working great for 5 plots, but now it keeps crashing after p1 table 1, saying it ran out of memory. I was looking at system monitor and RAM usage was not even at 25% when it crashed. I suspect that it is not able to use RAM that is currently used by cache, since cache is showing as 180gb right now. Any ideas on how to fix this? I ran sudo sh -c “sync; echo 3 > /proc/sys/vm/drop_caches” and saw the cache usage go down to zero, but the plotter still fails with the same error. I rebooted and it worked fine for another 9 plots and then ran into the same error

It might be running low on VRAM, did you monitor that? Try -S 3

That seems to have done the trick, thank you!

It’s a common issue with 8G cards + running a GUI, gets very close to the limit with the default -S 4

Hey Max,

Is there a way to tell if the gh farmer is running vs the regular chia farmer? I started the harvester with ./chia.bin however, when I look at farm status it still only seems to show plots from my original chia config and not my new compressed plots that I’ve added in the gh farmer config. Unless I’m just adding the plot directories wrong.

I’ve been using Flexfarmer for a while so a little rusty on the chia node commands.

I think time to refresh plots in latest release config is 2 hours. So it might be the reason if you creating plots fast in already added directory?

You can check it via ./chia.bin plots show

I just figured it out… you clearly stated in the readme that the node still uses the old config, I just missed it

1 Like

So far I’ve plotted on the server… and then installed the full HDD in the Chia client after 1-2 days.

If I have understood the Gigahorse plotter I need 2 GPU’s.
1 x PERMANENT in the client to complete the plots when the filter passes.
1 x GPU to generate about 70% of the plots.

correct ?
p.s. I only have 1GB network and it’s not getting any faster.

@Madmaxx
Chia itself would like to release a GPU plotter soon. if necessary, the vlt. not so fast with the plots. but that doesn’t matter anyway. I need about 10 minutes to move the plot to the hard drive. At 70% then only 7 minutes.
Accordingly, I could then also use an ancient graphics card in the plotting server.
BUT: Why should we take your system and pay (increased) fees. And then close software.

Did I miss something?
p.s. Yeah… you’re the first to put it out. But I’d be happy to wait another 1 or 6 months.

Halvening is next year. So far my plots are also 2% smaller, so the fee when compared to Chia’s compressed plots is negative for CPU, and just ~1% for GPU.

Another important factor is farming efficiency too, with GPU farmer you can simply make more profit (because you can run higher compression level for same power draw).

2 Likes

Halvening is one thing. But that says nothing about when chia itself offers gpu plotting.
And even if it’s only a year away (which I don’t think) we’ll all have to replot again.
I’d rather wait for the official gpu plotting.

Or have I overlooked something?

No, you might as well keep farming the old plots :wink:
There’s nothing to see here.

you’re dodging again.
Was hoping that you would have some arguments why one should use your gpu plotter.

chia itself has already announced one. Then you have the advantage that you can switch back and forth between the pools. And normal fees.

One argument would be: you are the first to bring it to market. any other benefits?

I assumed you are trolling.

Here’s the full list of features:

  • GPU plotting with 128G RAM (ie. partial RAM mode)
  • Multi-GPU plotting to make better use of RAM
  • CPU / GPU k33 + k34 plotting support
  • Higher compression levels C7 + C8 + C9
  • Plots are ~2% smaller than Chia’s at the same compute requirement
  • CPU & GPU farming support for k32 + k33 + k34 + k35
  • With GPU farmer you can get ~10% more rewards with same power draw (ideal case)
  • You can start plotting and farming now (instead of Soon™)

Chia is trying to get something out quick now, so you will see incremental improvements over time later, making you replot again and again until they’ve reached the same compression % and farming performance.

3 Likes

Define official, and your free to keep waiting and waiting like the battery commercial …