madMAx released GH 3.0 today. C29-C33 compression

Let @Ronski investigate a bit more his card temp, whether it is already throttling. Maybe that will give us a bit more info about the state of his harvester / GPU.

By the way, he already had 20 secs partials before we started this process, where his harvester is sitting on ~350 TiB raw. As you mentioned, he should be able to double the load. At that time, he had ~800 c31 plots, where his card should be able to pull about 4k (maybe around 80%) of those plots with the rest either c30 or c18 (i.e., maxing the GPU for 512 filter).

1 Like

no one bothers to share such a info ?

1 Like

You can roughly extrapolate it based on 3060 ti. RTX A5000 drops by more than half with every compression step, so assuming 3060 ti would have 12 GB VRAM, it would be close to 240 TiB. As non-ti is about 30% slower, you are looking at around 180 TiB raw.

Also, you can run ProofOfSpace (https://github.com/madMAx43v3r/chia-gigahorse/tree/master/chiapos#farming-benchmark) on one c31 to get possibly better numbers.

1 Like

I asked about RTX 3060 - 12 GB coz a RTX 3060 TI - has only 8 GB

1 Like

Therefore, I wrote that you can extrapolate it from an ideal 3060 ti (that would have 12 GB VRAM)

1 Like

I can tell you what a 3060 12GB can handle with regards to c31, but not c32, I knew testing would be pointless for my farm size of 323 TiB.

c31 on 3060 12GB

Partial Difficulty: 10000 (0.00555701 % chance)
Max Farm Size @512: 0.335287 PiB (physical)
Max Farm Size @256: 0.167644 PiB (physical)
Max Farm Size @128: 0.0838218 PiB (physical)
Average time to compute quality: 0.477824 sec
Maximum time to compute full proof: 0 sec

1 Like

Thank you!!
But also, I m going for C32.

What size farm?

20chrs

22 PiB split on 77 x 18TB computers

1 Like

Well you’ll need plenty of compute power, why 3060 though or do you just have plenty of them?

I have enough of 3060 and 3060 TI. this is my main problem. I dont have big GPU like 3090 or 4090

Hi, does anyone know the syntax for plotting on a remote server on ubuntu? Basically, it’s running plotter on different server and when plot is ready have it copy to another server when. I tried @//192.168.1.100\plot1 but it fails, thx in advance!

cuda_plot_kxx -n -1 -C 7 -t /mnt/ssd/ -d @REMOTE_HOST -c <contract_address> -f <farmer_key>

1 Like

You just specify @192.168.1.100 no extra back-/forward-slashes, no folder names. You need to run plot_sink on the remote side and specify plot folders on that remote server.

2 Likes

Because I began on windows(3950x, 128gb ddr4@3600) . Resisted the gpu craze and kept non compressed. My friend Ronski took the time to explain to me the whole gpu thing, helped me thru bladebit crashing and making bad plots then informed me of Max and gigahorse. So I tried windows plotting with a 3060ti. I could have used my cpu and gotten those times (Almost). Told me of his Linux experience. I have some Linux experience and wasn’t gonna swap that pc at that time. So I just bought a z840 loaded, thru a couple old sata ssds I had lying around and dual boot thru boot menu. Endeavor OS runs great with 512gb ddr@2400 and dual dual e5-2699v4 xeons. If it had pci-4 I’d be in heaven. The z840 is a plotter and that’s all. Slap 4 hard drives in, reformat the drives to ntfs, fire up gigahorse and green with envy and away we go. I may one day go to Linux for farming. (install rant.exe here) I’m not a winderz fan boy and find it a patched together mess. The whole pc software industry exists trying to make windows secure or run. Windows 10 = windows 7 with a bs layer over the top. I actually go to the trouble of editing the registry to get rid of that bs. This is a hobby and I just went the quickest route to get plots on the network. (rant.exe has stopped)

2 Likes

You’re the best, that did it, thank you!

2 Likes

Don’t know what brand you have. Repasting helps, fan speed helps, extra 120mm pointed at it helps. Increasing case intake fan speeds helps, making sure your ram contacts the heat sink helps, (I say this because I have ran into models with a heat “plate” that covers the ram with no contact to the actual heatsink). If it has a back plate I use thermal pads to “connect” the back of the pcb where the gpu and ram on soldered on to the backplate. After that I use one of these to give the back plate fins.https://www.amazon.com/Aluminum-Heatsinks-Motherboard-Conductive-Adhesive/dp/B09XN44MGX/ref=pd_aw_ci_mcx_mh_mcx_views_0
3090 while plotting usually runs @ 53c, vram around. 56c.
At 100% and 85% power limit it boosts to 2130mhz.

1 Like

I did re-pad my 3080 which made a huge difference, the 3060 is running in a rather warm environment, with plenty of air flow through the case.

It’s a ASUS DUAL Geforce RTX 3060 12GB V2 OC Edition 12GB

2 Likes

Having upped my difficulty to 2500 Saturday night, my EC has been steadily dropping, not had a partial for 5 hours now, and my EC is 200 TiB down.

Just changed my diff back to 1000, hate to think what it would be like at 20,000.

Edit. Got a partial within 15 minutes of changing back to Diff 1000

2 Likes

With high diff partials are random just like solo farming. Just need a bit more patience, it works fine :wink:

I’m running 400 TB effective at 5k diff now.

2 Likes

Ok, I’ll up it again when I’m fully replotted, and leave it a lot longer.

1 Like