New to Chia - Rig talk + is it even worth it?

Hey all,

Over the last few I have been playing around with chia plotting. I took one of my computers out of mathballs, an i7-2700k overclocked to 4.6ghz + 16gb ram + 6tb slow storage to mess around with chia. I added a 500gb WD Black SN750 NVME running through a PCIE to NVME adapter and got to plotting. Running one plot at a time because of the 500gb NVME I gave each plot all 8 cores, 8gb ram, and let it get to it. Average plotting time was about 5 hours per plot. I have now filled all 6tb of slow storage so I am at the point now of decided if I want to get serious with this or walk away.

By getting serious I mean I plan to run 3 separate systems, one desktop and 2 server. The desktop would likely be a 1st gen Threadripper 1950x and 64gb of ram. A guy is willing to sell me the system complete with AIO and an expensive case for $500USD. I figure that, whatever happens, I will be able to sell this system at a profit. With this cpu and ram + 4tb of NVME storage, I should be able to run 15 plots in parallel, and I imagine 30 per day.

The server systems would be something like Dell R720’s or some other 2u Dual Xeon E5-26xx v1 or v2 based older model server that I can pick up rather cheaply. My plan here would be to swap the CPU’s for e5-2667 (8 core, 16thread 3.3ghz 4.0ghz boost) but I will likely only run 1 cpu in each server to start. I may add enough ram to run a ramdisk but I am leaning away from that. On these I might run an NVME through a PCIE adapter but likely not. My plan for the first server is to throw in 4 used 800GB SAS SSD’s in Raid 0. I imagine in that system I should be able to run 12 plots in parallel and probably 20 per day. In the other system I would go slow storage - just for shits and giggles - I am thinking 6 x 600gb 10k or 15k SAS HDD in Raid 0. I imagine I could pull 14-15 plots per day from this.

Is it worth it:

I think with those rigs I could do a decent amount of plotting. Throwing in the second CPUS and throwing in more temp drives would obviously increase that. So lets say I can do a conservative 60 plots per day on these rigs, which is, roughly speaking, 6tb of plots per day - so I need to provide for 6tb of slow farm storage every day.

Now, based on calculators, as of today if I have 1000 plots I will make about $3.71/day. Obviously, 6tb of a day of storage is going to cost me more than $3.71/day. So lets say I plot 6000 plots - so 100 days of plotting before I sell off the plotting stuff. Well then I expect to make $22 per day or $8200/year. But to get those 6000 plots I need 600tb of farming storage. Assuming I can get good used storage for $12/TB that is a cost of $7200 in storage space. So, at the end of the year, after all the investment, not taking into account extra equipment and failures - just considering the price of farm storage, I would stand to make a profit of ~$1000.

But there will be equipment failures, we do have to pay for the plotting equipment, we do have to pay for electricity to run this all, and on top of that we have to consider the amount of time we put into this. And all of this, of course, does not take into account the growth in net space over a year which will significantly diminish monthly profitability over a year.

So my question is this: given that this is the case - assuming my calculations and assumptions are correct - what is the point in mining? I mean, the only reason to get into CHIA at this point is because you believe in the coin and believe that it will increase in value over time. This being the case, why the hell would you mine it and not just buy it? I mean, it seems to me that there is far less time and risk involved with just buying the damn coins.

Have you considered MadMax for plotting?

I am farming (approx 250Tib) and also have bought coins. If you believe in the project, then it is your decision as to which way you should go. At this moment in time, with XCH at approx £100, I would probably just buy the coins but I also like the fun of plotting/farming. That is why I have a dual strategy.

1 Like

Thanks for the reply…

If I go through with this all I do plan on looking into mad-max plotting. I just did not bother to learn it for the 6tb of space I have right now…

What got me interested in Chia plotting (not just Chia in general) is that its cpu plotted. I have been looking into playing around with servers and dual cpu system for a while. Years ago I was planning on building a Monero mining rig with a bunch of dual xeon servers but I did not get around to it and its certainly not profitable now.

It just seems to me that, even at $10 a TB for farm storage (which would be tough) it just is not profitable. On top of that, the rig building and plotting is what I find interesting but, in the end, its sort of the small side of the game. I guess I could run a pay service for plots once I am filled…

I have been on the same path as you - see post here: Proliant DL380 g6 Server slow plotting - #24 by teeboy - some of the G7 servers I talk about are much lower spec than you were looking at and really low price!

I have sold most of the servers now - my theory is for < £100 you can get a server that will do a plot per hour in memory and save to cheap disks. This is less money than a single SSD drive. Only use Madmax on the servers, then move the plots to a cheap farming PC.

I wouldn’t bother optimizing for plotting speed right now. Just get a standard nvme + 128gb of ram and a good last gen high thread count CPU and you’re good to go. If you’d rather invest in a farm without the plotter, you can also order plots online for ~$3/tb (nothing much) and download them directly.

If you’re looking for a short term profit, you’re better off not starting tho. It’s not a short term game, you’re going to profit over many years. Consider the following : the IPO should happen in the next ~12 months, the first halving is in about 2 years, electricity costs are non-existent (it’s a low opex business), the drives have a very very high life. Depends on how much capital you have and are willing to invest (it’s a high capex business). A good starting rig should let you plot ~$20-30k of drives per year.

Feel free to browse the forum and get a feel of Chia :slight_smile:

1 Like

Your plan could well work, but it is quite complicated and energy expensive, needing considerable space, time, effort, and upkeep. If, as you mention, you enjoy playing around with older hdw, it will be fun in its own right. Your profit calcs may be overly optimistic, but it’s a nice thought we all hope for.

Best to 1st decide what your intent actually is. Do you more want to work with the hardware you outline, or do you simply want to plot and farm Chia?

Aurelius is spot on IMO. If getting involved with Chia is the choice, then I recommend ‘Keep It Simple’. One current gen multi-core MadMax plotter & small farmer, correctly configured, can do what you envision (60 p/d & farm 1000’s of plots) with minimal energy, space, and fuss. Chia is a long game, so I like to think of Chia hdw as an appliance, not a vocation, and let that drive use/buying decisions.

Lots of war stories on the forum, take heed and learn from others efforts, then decide. Good luck, whatever you decide, enjoy your journey.

Your i7 2700K, with a good NVMe 500 GB on PCIe slot, should be able to do 2 hours a plot by Mad Max. 8GB ram is more than enough. That is 12 plots a day.

You will not see any profit in short term. There isn’t even a marketplace where you could cash out XCH without paying huge commissions.

For long term farming, strive for efficiency: correct use of hardware, low operating costs.

You can have fun building a plotter or many plotters, doesn’t matter how inefficient or ugly they are - from reusing any components at hand to buying top tier equipment. Depends on your goals: do you want to plot as many plots per day as possible? Or you prefer a slow/cheap filling up available HDDs.

But don’t skip on quality hardware for the farmer. Ideally it should be server-grade equipment with JBOD appliances added for scaling. Running full node requires a good SSD and decent CPU, especially when network is under DoS attack or otherwise strained. Cheap/shucked HDDs may not serve well. If you go with the highest capacity drives (18TB) then make sure they come with the 5 year warranty.

We’ll see how those farmers who built a rat’s nest of USB drives going to hold up in 2-3 years.

1 Like

When you say they do a plot per hour in memory are you referring to a ram disk or with a simple spinning drive?

So if I understand this correctly - you would run slower servers and slow arrays but with high thread count, ram and storage you can do enough slow plots to equal 1 per hour?

The only draw back to my current setup is that the pice adapter runs at 4x and its Gen 2 so the max read write I get is like +/- 1100 mbps.

I just picked up a good used 4tb hdd for $40. I will throw that in and do some mad-max plotting. I looked at mad-max last week but couldn’t really understand how to set it up.

I just have a really old Server with 2 x Xeon Hex Core chips. 144GB of RAM setup as a RAMdisk. Use Madmax to create 1 plot at a time - and it will pop out 1 plot per hour. I have been selling these servers on ebay for £80 or 90 - the plots are stored on the SAS disks at the end, so then you copy them off via network or USB to your farming PC. Linux is much better at using all of the cores than Windows.

I am not endorsing this vendor, but if you want to go high-end, you could build a threadripper with 256GB RAM:

The large RAM capacity with threadripper is good for madmax.

So the Threadripper system is just something that popped up independent of My Chia goals - it is an awesome deal on an awesome system…that said 16 cores and 32 threads plus lots of ram and pice lanes seems to make it an awesome chia platform.

The server systems would be bough specifically for Chia ie. it is the excuse to buy them. I was recently offered an HP dl360p g8 with 2xe5-2620v1 cpus and 96gb ram for $200. I may buy it but it is 1u and I am really hoping for 2u servers so I can pack them full of drives. My hunch is that for $200 I can’t go wrong and with a bunch of cheap used enterprise ssd for T1 and hdd for farming. I think a system like this would allow me to plot a good number of plots without over running my storage space.

I do have an x570 board lying around and have been contemplating grabbing a used 5950x, 5900x, 3950x or 3900x or Xt for it. I got the $250 board for $100 so I have been contemplating building a high end rig with whatever good used parts I can find. For pure a chia purposes this system would be more powerful than I need and I likely would not be able to keep up with the storage space for the plots it could create - but I also don’t have to run it at full blast…

I am not orientated towards the short term with chia. My plan was to mine at least 15-20 coins and then sitting on the coins to see where they go. But when I see the relatively high cost of storage and the relatively low value of Chia, and the fact that it’s not the easiest coin to move if you decide to get out, it just seems to me that buying the coin is a better option.

There is also Lenovo (formerly IBM),

With either vendor, upgrading RAM or CPU yourself will be cheaper.

So are these like x5650 or similar bases Xeon systems? How, exactly do you use the RAM disk? I assume it’s used as a T2 drive but I’m not sure how that works…does it just offload some of the work from the ssd/nvme?

I can get some pretty good deals on 800gb sas drives with like 17 pb endurance so no worry that they will burn out. 6 of these divided into 2 arrays of 3 drives in Raid 0 should provide a relatively fast T1 storage of 4.8tb combined.

The primary bottleneck during plotting is during the External Sorting. This is an important feature, because it prevents ASIC attacks, resulting in the strength of the Proof of Space design.

Large amounts of RAM improve your External Sorting speed.

1 Like

Hindsight is always 20/20 :slight_smile: That may be the best plan (perhaps even a great plan lol). Twenty XCH @current $156 = $3120, and will exceed what you can reasonably expect to get from a farm of several 1000 plots in quite some time (1 or more years) that can cost times more than that. Maybe not so much fun and any learning experience, but far cheaper way to do investing in XCH.

So I learned madmax and got to work.

Running all cores, T1 and 2 to nvme, and 256 buckets I am getting between 90 and 100 minute plots.

I’ve set madmax to run infinite to a 4tb drive and I will check on it tomorrow. I assume I should be able to do 14-16 plots per day.

Not bad for a cpu I bought at like 40% off in 2011!

Edit: so I checked on it just now and it had either crashed, rebooted or in some other way restarted after completing 5 plots. Got it up and running again will see what it does.

I am curious to how my old amd A10-7800 system would handle chia. Slower less core cpu but it’s pice 3 instead of 2 so the NVME should be considerably faster.


I have a small farm to be interested and to consider making it larger, I bought coin instead.

1 Like

Even old machines can plot. If you’re not in a hurry, gather some old machines and run plotters and harvesters on them!