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

This OC’d I7-2700k system is doing quite well. I have it running mad-max stable now at about 95 minutes per plot or about 15 per day. The I7-2600/2700 are amazing processors. Until Ryzen came out with its bagillion cores, and Intel was still stuck with 4 cores/8 threads, the old I7 was only about 30% behind nearly 10 years after it was made. But that speaks more to the problem of lack of competition in the CPU market that allowd Intel to coast.

Tomorrow I will try an old I7-920 18gb ram system I have lying about. If I can get 11+ plots per day on it I will probably run both systems at once.

I tried running Mad-Max on the AMD A10-7800 system. Ya…no. Way too slow. Whereas the I7 does P1 Table 1 in like 47 seconds, the AMD system took 2 minutes and it got worse from there.

If the CPU is the bottleneck, try running multiple madmax with lower cores each, and simply plot to multiple HDDs in parallel.

I have been so tempted lately to make a Chia mining rig. I have an X570 board lying around and have been contemplating throwing in it a Ryzen 9 39000xt, 5900x, or a Ryzen 7 5800x. I was also offered for $400 an HP Proliant server with 2xXeon 2670 and 192gb of ram. The server would be awesome because of the ram-disk capability.

But in the end I have decided, at this point, to stick with the single i7-2700k machine. And here is why:

With this computer I can plot about 100 plots per week +/- 10% if I run it continuously. The problem is that I am not willing to continuously buy 10tb/week of new HDDs. So I am forced to look at the used market - which I am fine with. But here, in the used market, finding 10tb/week of used storage can be difficult unless one is willing to buy 10x1tb drives, which I am not.

This week I got lucky…I managed to pick up 8tb (2x4tb) of internal HDD + 18TB (3x6tb) of external drives for $370 or $14/tb (This is Canadian dollars mind you, so roughly $11.30 USD/tb.

I got lucky this week but it wont happen every week. And if I spend the money to buy a new PC that can do 40+ plots per day, or spend money on he server that might do a similar number, I will just more quickly run into the HDD problem.

Note: the server has no disks and is setup for the SFF 2.5" drives. If it were setup for 12 3.5" drives I would have bought it without hesitation. But large capacity SFF drives are just too expensive.

1 Like

Yes, plotting speed doesn’t really matter if you don’t have drives to fill, I think you made the right decision.

There has been no mention of electricity cost here at all. Is this because the expectation is that Chia will rise in value considerably and therefore it is irrelevant? I am hoping this too, but also have an eye on what I have spent so far and what I have farmed.

I am now using a Dell R620 (dual E5-2670) to plot, with 384GB ram (though only 128GB plus whatever the operating system needs are required as long as you have a modestly fast temp 1 drive - doesn’t need to be an SSD just not the slowest spinning disk available) and producing a K32 plot in 30 minutes. However, running this machine a reasonable amount results in about 200 kWh per month in power consumption.

Add to that what the farmer machines require and it’s pretty close to being just about break-even at this stage, not including the cost of any equipment, just on electricity.

Of course, if the price of Chia one day hits the heady heights that bitcoin or Ethereum have reached then this will all be irrelevant.

The waste heat warms my house, so zero spent on natural gas.

The amount of electricity used over the lifespan of the plot makes this basically a non-issue. You can do the maths and see the breakdown of costs for yourself : most of the cost of farming chia is the purchase of the hard drive.

Your price per kWh must be lower than mine. Electricity is my main cost. My equipment is all second hand and can be sold for what I bought it for.

I am paying £0.19/kWh.

That is a good point, but it is not something I am really willing to get too much into the weeds over yet. But at this point, with the i7 right, I am getting about 100 plots per week. Lets say I plan to plot 2000 plots, it will take me 20 weeks to get there. After that I will not have to worry about the electricity cost of the plotting rig. That said, I am pulling about 150 watts from the wall or around there. So, running 24 hours a day that is 3600 watts per day or 3.6 kwh/day. Over a week that is ~25.2 kwh which, at an average of $.13.kwh here is $3.30/wk. So, over the 20 weeks I will likely run this machine it will cost me $66 in electricity.

Considering I already have the rig and all I have added is a $75 WD Black NVME without HDD costs my total expenditure with electricity is ~$140 Canadian. Based on chia profitability calculators, once I get 2000 plots I have a 53% chance of winning per month, an EV of 1 xch per month. At $200 a coin that is great - theoretically all the rig expenses would be covered in 1 month notwithstanding I will have to replace the NVME once or twice.

The point is, there is really no comparison between the impact of plotting rig costs + electricity and the cost of the HDD needed to farm the plots. It it is surely the cost of HDDs that holds us back. I mean, one 18tb here goes for about $550 Canadian after tax. To get 2000 plots I would need 11 or 12 of these - lets just call it $6000. Even if I can get a huge drive used for $15/tb, which I cannot, it is still $3000 in drives alone.

I lived in the UK for 3 years and it seems to be that the price of energy there is just significantly higher. I was in Oxford and was paying I believe about £.23/kwh on average. Converted to Canadian dollars that is ~$.40/kwh. Here in Toronto I am paying $.13 on average when you consider peak vs off peak rates.

I’ve been getting drives cheap maybe £8/TB buying some used untested drives and taking a chance most have been okay.

So I can sell them again for more possi ly as they will be tested.

In my business model hard drives aren’t a huge cost, they’re a cost and I’ve spent several thousand pounds on them. The only sunk cost is the electricity.

Planning to get solar and only plot when I am in surplus.

I calculated estimated costs of running a 504TB farm using 18TB drives vs 8TB drives

Assuming both 18 and 8TB drives use 7W/h or 0.007 kW/h
Assuming price $.16 per kWh

You need 28x 18TB disks or 63x 8TB disks to achieve same capacity.

For 8TB disks:
Usage 441 W/h or 10.584 kWh/day or 3863.16 kWh/year
Costs $12/week or $620/year

For 18TB disks:
Usage 196W/h or 4.7 kWh/day or 1716.96 kWh/year
Costs $5.26/week or $275/year

Besides that for smaller drives you will need more capex costs: jbod appliances, cabling, HBA. Plus opex overhead (+ energy consumption for multiple jbods versus 1-2)

See this link for easier comparison :

Off topic but damnnit it’s my topic so here I go…

What are the prospects for selling plots? Like, if I built a mining rig how well would I do selling plots to others once my drives are full?

I am still contemplating buying the server. Guy is offering it to me for $350 Canadian and it has 192 gigs of ram. I mean I could likely sell the ram alone for more than I would pay for it! I know it wouldn’t be the fastest plotter but with 2 e5-2670s and a RAM disk I’m sure it would be ok. Just wonder if I could sell plots with it after…

Also, if I were to run 2 virtual machines on the server, each with 1/2 the ram and 1 full cpu, would I be able to run to instances of MadMax? I was thinking if I bought it I could max out the ram and run 2 instances of madmax each with a just ram as the temp storage.

My impression is that plot-selling businesses haven’t been very successful so far. The biggest challenges are logistics and trust. Plots are very large, unfeasible to move over the Internet for the vast majority of people, and you need to supply your keys for them to be able to plot on your behalf. Couriering hard disks around isn’t much better either - it’s reasonably expensive and your goods have a decent chance of getting damaged in transit.

There’s also a good amount of competition already, and most will be a lot more established, have websites, prior customers, etc. It’s also my feeling that if you want to plot at scale, then the costs make it so that you’re better off buying your own plotter. If you don’t want to scale, chances are you don’t want to pay anything either. It’s a tricky business.

I’m not trying to dissuade you, I think that’s a good deal if you need a plotter for yourself, or if you just want a new toy to experiment with. I just wouldn’t buy it in hopes that it will make me money by plotting, I don’t think that’s likely.

That’s a decent chunk of RAM, though it’s probably DDR3 so not very resellable. You need a bit more of it if you want to use only RAM as temp storage (check out requirements for MadMax or Bladebit). It is enough RAM for MadMax to only have to do 25% of writes outside of RAM though, so still a positive.

I hope this helps.

1 Like

It helps for sure,

Nice to have a place where you can sort of spitball ideas and not get talked down to by people in the know!

I wouldn’t buy a plotting rig to sell plots I am just thinking about once I am done plotting or down time while saving up for more drives.

I imagine I would need about 300gb of ram for pure RAM disk plalotting and 600gb to do all ramdisk plotting on a 2 VMs. Considering it’s ddr3 server ram it is certainly doable. And for the cost it would be similar to buying NVME drives that will burn out!

1 Like

Arriving late (and the helpful crowd here :+1:) gives you the advantage of sharing others experience and their best tools.

If I were starting a Chia Farm now and had the resources to spare I would love to do RAM disk plotting.

Good luck and happy plotting!

1 Like

The problem with selling plots is the bandwidth necessary to transfer the files : depending on the speed of your connection and of your plotter, you will probably saturate your home internet with ~1 plotter.

Also as was said in previous answer, that amount of ram is not enough for pure ram plotting at k32, and you’re likely going to need more ram space for high k counts.

I would advise you to invest small amounts regularly over time and not aim for the “best solution” as this is a fast changing space. For example, you can now see drive companies pushing new high endurance SSD models which can bring the drive-cost of plotting down to ~$0.15/tb.

Thanks for the advice.

At this point I am finishing up a Ryzen 7-5800x, 64gb ram build. My plan is to plot with that until I reach hdd capacity then move onto some other cpu mined coin while farming chia. I finally broke $1/week in my pool in chia profits…yay?

Anyhow, currently I am farming 90 plots and have a total HDD capacity of 34tb. I should have enough space for about 300 total plots. At this point I am not investing in any more hdd unless it’s 6tb+ and I get it for under $10 Canadian per tb.

I did alright on the used market lately though. Picked up 5x6tb drives for $400 or about $13/tb.

2 Likes

Alright, so I finally go around to building the Ryzen 7 system I’ve had lying around - still waiting on the case so as of now its running on my dresser on the mb box!

Anyhow, this is quite the little Chia performer. Here are the specs:

Ryzen 7 5800x
Aorus Elite Wifi MB
16gb DDR4 3200 clocked to 3600mhz
500gb WD SN750 NVME (1 drive for both T1 and T2)
128gb SSD for OS
WD Blue 6tb HDD for Farming

Right out of the gate, with just a small overclock to the memory clock, I am getting ~45 minute plots, or about 32 plots per day. Left running for 40 hours I should be able to fill the entire 6tb drive. Sure, there are faster rigs, but this is pretty simple with 16gb ram, 1x500gb nvme, and 1 hdd.

Currently I am farming 90 plots. I have enough HDD space now to farm 270 plots. Once I’ve filled all the farming drives I will likely move on mine some other coin until I find an amazing deal on a huge hdd.

2 Likes