Any advice/experience in "low-end" GPU plotting with old-fashionned SSD?

currently 70TiB all are gigahorse c8 k32 plots for Chia, my difficulty is 20 and my lookup times are about 15-20 seconds. this is my flexpool : i am using flexfarmer, but in my experience that doesnt make a difference to lookup times because even with a ping of a second it doesnt make a difference when the k2200 is making lookup times 15seconds

1 Like

I am solo farming with 600 TB with Ryzen 2200G + 16 GB RAM (Very stable and good rewards)

I have another PCs too
Ryzen 3600 + 32 GB RAM + GT 710 2 GB + 2 TB NVME
Ryzen 3600x + 32 GB RAM + RTX 2060 6GB + 2 TB NVME

My options:

  1. No change. No need powerfull pc for farming.
  2. I will spend money to server that supports 128/256 RAM and earn 20/30% more and I will give 3% to madmax. This is really stupid.
  3. Compressed CPU plots. i dont know maybe level C4

What do you think?

1 Like

Why don’t you do nothing new, continue as you have? This is super easy for you. Dabble in C4 or whatever plots as you desire over time. Why stress & strain?

1 Like

Hi everyone.

I finally did it. I bougth my Z440, not as a plotter, but as my new farmer (which use to be an old mining rig).

I use Windows “Storage spaces” to have bunches of 4 disks (linked to USB hubs).
Each hub goes into a Z440 USB3 port (added 5 of them with an internal Amazon hub powered by a SATA power connector in the Z440).
I use 10 ports, that makes my 40 USB disks fit.
They are mounted in two home-printed racks, and powered by 2 350w power supplies 12v rails (no more ugly AC/DC converters spaghetti, did some soldering here…).

The z440 is W10, but runs WSL2 with Ubuntu to flex-farm the compressed ans old plots.
I mounted the GTX 1060 for farming purpose (at the end it should ~440TB of C8 compressed plots).
Plot_sink gets the plots from other machines and distibutes them.
An home made script delete one old uncompressed plot on each 4drives-array if its free space gets under 200Go.

I got my 2 olds CS3030 1To re-purposed for good’ol’fashion plotting… Plan to expose their TBW after that. My 2 plotters produce 1 plot every 55 minutes : did my math, getting a 3060 Ti + RAM does not really cover the cost of getting the RAM : I stay on CPU compressed plotting, and should be done in ~3 months.

My 3060Ti stays in the gaming computer.

Thanks for your advices, and thanks to Digital SpacePort, Max & Flexpool team for their respective great work.

1 Like

For those that don’t know NVidia workstation cards, the K2200 is pretty much GTX 750 ti with slightly lower clock speed but has a lot more RAM available.
Don’t try to plot on it, plenty for most smaller farms.

2 Likes

avoid compressed plots for GPU farming, especially if you pay the usual 0.3+ €/kWh

C4 compression study will be available in two months.

I used to plot ~6min with MP600 2TB, Ryzen 9 3950X @ 4GHz, 128GB DDR4-2133MHz, 3060Ti. Your plot price will be around 0.05$/plot. Complete replot ~200$.

If you have no 128GB RAM, which will set you back ~600$ (unsure about price now). You can buy used server for same money. I just bought one that plots ~200s plots with 3060Ti.

Do your math…many kids jump the gun, without even knowing their profitability…talking about “fair” price 300$, predicting the future.

For me, official plots still make more money than C8 MMX plots. There is something funky with FlexPool farming…unsure, whether to rather wait for the official release.

Pick your poison.

My poison has already been picked.
I farm C8 on 1060 6GB.
I plot on my usual “madmax” CPU, no GPU used after reflection.

It costs literally 0 to me.
But I get more ans more rewards…

1 Like

how do you plot c8 on CPU?

chia_plot.exe -n -1 -k 32 -C 8 -x 8444 -r 5 -u 256 -v 256 -d @192.168.1.28 -t K:\ -2 K:\ -s K:\ -f **** -c ****

192.168.1.28 is the farmer with chia plot sink running…

how come i was mistakenly thought only c0/c1 is supported with cpu plotting … hmm

thanks!

2 Likes

I guess you were confused with the farming limitations of CPUs?
Anyway, it’s still plotting here, slowly but at a ridiculous cost of 3 threads running continuously on each of my 2 Ryzen 3900X.

I read your thread with - how to put that - “interest”, but agrees totally with the answer Max gave you.
Max delivers, thanks to him & Flexpool team, I’m in, and am happy to pay my fees.

At least, my farm is as profitable on Flexpool as it should be according to my number of plots with the hindsight of two months of feedback… :slight_smile:

1 Like

you can be right or do it right. choice is yours. sheeple stick together - after all, shared truth is the only truth right ROFL

Just few hundreds years ago, sheeple believed Earth is flat…and they are still here despite the 10y olds know Earth is round ROFL

You mean you didn’t see this
image

1 Like

or do you recall that global IQ test between 2019-2021 where we had the deadly disease that killed brutal amount of <1% of population.

over the 70% of population jumped the gun, without asking why, and got jabbed by experimental gene altering crap that killing people en masse now ROFL

Everybody whines about evil Russians, but they cured the disease in less a week just by saving Ukraine ROFL

and the next year when electricity bills comes in, the same expert will say “I knew from beginning GPU farming was crap because that guy with ‘wrong’ math told me so.” ROFL

I bet all my XCH on it ROFL

Who is in? Let’s call it Lobotomite pool - winner takes it all.


I did the same thing lol. It’s saying those compression levels are only appropriate for CPU, not that they are the only CPU farmable compression levels available. It’s not super clear where the CPU vs GPU line is probably because it’s variable and gear dependent.

3 Likes

I read it that ALL levels are plotable but some are practically farmable only by cpu or gpu

For instance, even if you could gpu farm c1, it’s not profitable

1 Like

That threw me off too, conceivably we could have ignored GPU plotting altogether and still reaped the rewards assuming the farmer was up to it (GPU or beefy CPU). Especially for small farms, it’s worth doing the math on what we really need!

Yes exactly, you’d need more than 100pib to make it worth the idle power of the GPU. It’s “CPU only” because CPU is enough.

1 Like