Gigahorse Compressed GPU Plot Client Released! Start Plotting Now!

MMX has released his client that allows for compressed plotting and farming at:
Link

I’ve written a guide based on his September release so some information may be out of date. FlexFarmer support will be coming.

Guide link

Gigahorse works with any pool but farmers may have issues farming them on other pools besides Flexpool due to the changes that need to be made to properly support them.

5 Likes

not that i don’t like flexpool but is there currently a list of pools that support the format of max?

I know flexpool has supported max, but it would be nice to have a choice and not be “forced” :wink:

have a nice day and happy replot :slight_smile:

2 Likes

All pools that use the official pool protocol.

But you need to use my Gigahorse binaries to farm, node + farmer + harvester.

4 Likes

ya it was to be assumed :smiley:

I’m looking forward to the weekend, finally rebuild everything again that ran now already 1 year through without that I touched it, thanks for that :slight_smile:

From what I understand for compressed plots pools will need to implement a few things including the variable difficulty system we have so farmers can set their own difficulty. The problem is most pools run the basic software from Chia that wasn’t meant to be used commercially. So their backend can’t handle too many partials. For instance our top farmer has submitted 950k partials in the past 24 hours while for most pools they keep each farmer under 500 per 24 hours.

Allowing farmers to choose diff can result in their backends crashing if the diff is too low. So to support MM plots pools will need to do their own backend software like us or risk crashing. That’ll take them weeks at a minimum.

What is the rough estimate on being able to use Flexfarmer to farm MadMax compressed plots?

Will we only need MM compressed plots and Flexfarmer to farm them?

Thanks!

It just came out! Give us some time to try it ourselves.

That’s the plan

1 Like

Glad to see your software is finally released.

Do you have any audit planned?

Unless the auditor compiles and releases the binaries, I don’t see how that would help…

Anything special I need to do to enable pooling? chia farm summary shows as synced and farming, but partials are not showing up in SpacePool for an hour now. Not sure if it matters, but right now I don’t have any compressed plots, they are all just regular nft plots. Thanks in advance.

I have to admit I didn’t test farming with regular plots, can you check the log file and see if there is any errors / warnings ?

EDIT: Did you stop all the previous running chia processes via chia stop all -d, and then use a new terminal to start gigahorse?

EDIT: And you’re running gigahorse node + farmer + harvester? no official chia node + farmer with gigahorse remote harvester (although that will throw errors)?

1 Like

Did any one try it with a small ram configuration of 16GB? 32GB? is any difference?

@madMAx43v3r On the main readme page, the first table refers to k32 plots (with respect to PiB / GPU). Can we expect the same GPU farming capabilities for k31 and k33 plots?

This might be a stupid question but is there a difference with “compressed” plots created by the gpu vs one done by the cpu. Can we plot via GPU then handover the work to the cpu? Taking care of course to use the correct compression level for the machine so that the cpu can handle it for the number of plots.

The reason I ask is that I run multiple sub PB farms for different groups (colleagues, friends etc). I plan to replot everything with one GPU and then sneaker net them to their harvesters.

yes you can, consult the tables here:

I would not go higher than 4 if you want to farm CPU only

1 Like

I saw that table alright and I actually think I’ll try to go for C5. They are fairly powerful modern cpus with plenty of ram so shouldn’t have any problems.

If I understand then correctly, there is no difference in the plot architecture between cpu and gpu plotted plots and it just depends on the power of the cpu (in this case) whether it can keep up to fill in the missing data.

Good to know, thank you.

Do not forget that, the higher you go in compression, the less energy efficient CPU farming becomes.
If power cost is of no concern then ignore this comment :slight_smile:

True, I’ll have to calculate that as well. It’s not of no concern but electricity here is not as expensive as some other places.

Even at full load 24/7 the electricity cost per month would only jump a little. Ill have to calculate whether a lower compression would give a better return however. Very valid point to make, thank you.

Didn’t see anything in the log file. I made sure to stop all previous running chia processes and even restarted my computer just to be sure. Yes, I’m running the gigahorse node + farmer + harvester. I guess I’ll just farm compressed plots while I replace the old uncompressed ones.

Yes, it’s roughly the same, higher K size is a bit less.

The plots are the same yes.

C5 is still fine for CPU, the levels are the same as NoSSD.

C6 is 4x times more power than C5 on CPU, so that’s where CPU start to become inefficient.

Let me know if it works for compressed plots. I just tested uncompressed plots myself, and it works… even with remote harvester.

2 Likes