Experiment: Build a (Profitable) Ethereum Mining Rig From Spare Parts

[TJ Hooker]

Spoiler

Eth mining is memory-bound, and needs high bandwidth, low latency VRAM. Integrated graphics using system memory as VRAM would likely perform very poorly. Also, you would need to allocate > 2GB of system RAM as VRAM, so the machines would probably need to have >4 GB of system memory.

I was thinking along the lines of schools/universities where computer labs are running all day anyway. Like a background task kind of thing. Generate extra revenue for the school.

If you're paying for electricity for these machines, I'm skeptical it'd be very profitable if at all.

Now you're getting into misappropriation of equipment. It would be bound to, at the very least, ban the user from using computer resources.

He was referring to the institution itself using its machines for mining. Not some student creating a mining botnet out of university computers.

I think I remember someone on Reddit saying that their school actually using some of their HPC equipment for mining when not in use as a way to make a bit of extra cash.

 

[extremepcs1]

Now you're getting into misappropriation of equipment. It would be bound to, at the very least, ban the user from using computer resources.

Sorry for the misunderstanding. I meant something that is sanctioned by the school and the revenue goes to the school.

 

[shrapnel_indie]

Now you're getting into misappropriation of equipment. It would be bound to, at the very least, ban the user from using computer resources.

Sorry for the misunderstanding. I meant something that is sanctioned by the school and the revenue goes to the school.

Okay, that's a different story. That is up to the schoo, school board, etc, obviously. They should also take into account their costs to operate their farm. Heat may be less of a factor after-hours, but it still can't be ruled out altogether.

 

[Ncogneto]

Schools and Business's could take advantage of the lower kw/h power costs provided to them. Heat can be offset by location (northern climate/ underground )

 

[kcarbotte]

Believe it or not, the Power Color RX 480, which is otherwise a very efficient mining card, dumps more heat into the room than any other GPU in this roundup. After a little more than 10 minutes of mining, the card reaches nearly 80°C.

That's not how it works...
There is not a direct relationship between GPU temperature and heat output (it also depends on the cooling solution). There is a direct relationship between power draw and heat output (they're equal). R9 390(X) are going to heat up your room significantly more than a 480.

You're 100% correct. I was not thinking clearly when I made that judgment.
We've amended the article. Unfortunately, it's too late to go back to change the rig. The RX480 is now in use in another project.

 

[kcarbotte]

Err.....trying to factor in card temperature is an utter waste of time. It doesnt matter if a card is at 10 degrees, 50 degrees or 100 degress. Thats just a factor of how well a heat sink can remove the power used by the gpu. It doesnt factor into cost calculations.

The only thing that matters is how much power it uses.

You do not spend a single cent extra if a card is at 100 degrees or 50 degrees. If it uses 200 watts it uses 200 watts. And if you are cooling your home instead of dumping the air outside, then it costs the same to cool a card at 50deg or 100deg as long as either case uses 200 watts. (note i know that resistance is a function of temperature, and a higher temp will draw more watts, but as long as you measure the watts, it doesnt matter. What the card draws at load is what it draws at load.)

It does matter for noise, since its likely the fan is spinning faster and thus noisier at higher temps.(which again uses a tiny bit more power to spin a fan faster, but again thats arleady accounted for in the load power draw).

And it would matter if you have to apply external cooling to cool your rig when mining. But this is also a function of total power draw, not the temp of the card. It a power density thing of trying to cram 1000 watts in a small space.

-------------------

Beyond that. watts/hash is not the only metric. It pretty much is for a huge mining farm. As well as the largest factor if you are power limited at your location. But if you are building 1 rig with spare parts id say its much more important to consider total mining hash rate. That is as long as the extra hash rate does not end up earning you less.

As an example: Option A gets 100Mh and earns $6 in coin for $3 in power = $3 profit, 33.33Mh/$. Option B gets 110Mh and earns $6.6 in coin at a cost of $3.45 in power = 3.15 profit, 31.88MH/$. Option C gets 120mh and earns you $7.20 in coin at a cost of 4.35 in power = 2.85 profit ,27.59Mh/$.

Option A is the most efficient. Option B is 4.5% less efficient, and yet it still earns you 5% more profit. Option C is 20% faster, but 17% less efficient, and earns you 5% less profit. Option B is the winner.

And thats not even counting the fixed cost of the base system. The cpu uses power, the memory uses power, the mobo uses power, the storage uses power. There is a base cost for having the system on.

Lets use the same example above, and add in 75 watts of base system power, and say we were talking about a days earnings, but we werent counting base system power. At .13c/khw(roughly us national average) that adds 24 cents a day in costs. That means Option A earns $6 for 3.24 in power = $2.76 profit, option B earns $6.6 for $3.69 in power = $2.91 in profit. Option C earns $7.20 for $4.59 in power = $2.61 in profit.

Option B is still the best and looks a tiny bit better, now earning 5.5% more profit on 3.5% less efficiency, option C is still the worst.

Efficiency matters, but just looking at $/watt on a gpu is not enough.

A good efficient power supply would be absolutely critical on a mining rig as well. If you are doing a 24/7 1000 watt load, at .15cents/kwh then having a power supply that is 5% more efficient would save you $65 in power/year If you tripple power cost, then it saves you up near $200 per year. On a 24/7 high power draw, the more expensive/efficient power supplies start to pay for themselves quickly with a big 24/7 load.


Note: I do not run a mining rig. I've only dabbled with using my current gaming gpu with mining when im not playing games. For me, base system cost is a rather large percentage of power draw using just 1 card(tho using a 8 core ryzen cpu also to mine pays for the base system cost, as well as the extra cpu power, and a lil bit of profit). And the platnium power supply i upgraded to recently is saving >10% on power compared to the 85+ rated one from 8 years ago. (Note i did not upgrade the power supply for mining, i did it before i even tried mining, old ps was insufficient, and i only really got the platinum one because it has a 10yr warrenty. With mining while idle however, it has turned out to be a really good decision)

This is all excellent insight. I see where you're going with the fixed base cost, but in this case, I don't think that matters as much. This PC's only function is as a dedicated miner. It would not be running if it weren't mining, so the fixed base cost is still a factor in the profitability of this rig. If you were to use your main gaming PC for mining, then you could discount the fixed costs of the base equipment.

However, I wasn't trying to suggest that it would cost you more money (other than potentially in AC cost, which isn't easy to calculate without a controlled environment) to run hotter cards. Heat output is a factor for comfort in the household. As is the audio level.

That said, the methodology of selecting the hottest cards was admittedly flawed. We incorrectly equated GPU operating temperature with heat output into the room. We've amended the article to address that issue.

 

[kcarbotte]

I was wrong, I forgot that heat pumps operate on a different principle than heat engines. For moderately efficient AC systems of (COP 4), the theoretical power required to run the cooling is in the vicinity of 25% of the original heat, though realistically it will be worse because of duct losses and other inefficiencies. So if you're using 1000W on computer power, your AC will consume another 300W or so, less if you have an above-average AC system.

I don't have an effective way to measure the power consumption of the AC system. I haven't lived in this home for long, so I don't have an idea of the average cost to operate the unit without the miner running. I also don't have an accurate way to measure how much of my overall power consumption comes from the AC. I have a central air system, which doubles as a furnace and AC, and is wired directly into the electrical panel.

The climate around here is mild this summer, so I've had the AC off and the windows open about half the time the miner has been running.
When I get my power bill, I'll have a better idea. I know how much the miner should cost.

 

[gnyff]

Probably should google this a bit more before posting, sorry... ;-) But the world is /actually/ using resources (time, parts, and electricity) of "mining" a (unbacked, but I guess most are ;o) crypto currency?!? No useful work is actually done (?); the machines are just calculating random stuff made sufficiently difficult to match the hardware of today? I hope I'm missing something here - but I say use the hardware/electricity for fun or work and leave the waste of resources to....hm... can't really think of a "purpose" that not either useful or fun... [2 degrees target long gone! And US out of Paris?!]

 

[kcarbotte]

Just a hypothetical, but say someone had (legal) remote access to 5,000+ normal desktop computers. If he/she were to install the software on them, could it be profitable? On-board Intel graphics type machines.

As far as I know, crypto mining requires dedicated VRAM and would not work on shared system memory. Most onboard video solutions can't access more than 2GB of system memory anyway.
I don't know for sure, but I'm going to go ahead and say it won't work.

 

[kcarbotte]

Great article, wish you would take it a step further and play with tuning the system to see how significant of a change you might be able to dial in. Also, In regards to heat, maybe it makes sense not to mine during the summer cooling months, but run it 7-9 months out of the year were the heat it produced would not be wasted completely? I understand it's not the most efficient way to heat a house, but, at least your not doubling down on costs.

I intentionally left the GPUs stock to see how well they would perform for a month.
I'm not done with this project, and I fully intend to dive into tuning and tweaking to maximize performance.

as for seasonal mining - you could absolutely do that, but don't expect Ethereum mining to be around this coming winter. The difficulty continues to rise, and proof of stake is coming (soon?), which will render miners useless for Ethereum.

There will undoubtedly be other currencies to mine this winter, though.

 

[USAFRet]

Just a hypothetical, but say someone had (legal) remote access to 5,000+ normal desktop computers. If he/she were to install the software on them, could it be profitable? On-board Intel graphics type machines.

Then you compete with these guys:

 

[helopilot]

Great article Kevin. I liked the way you got right down to business and approached the topic. It also shows that you have a great talent for figuring out a practical approach to solving (any) a complex problem. No one has asked how many man hours you put into this project? I'm guessing that you you would probably make more money if you'd spend the equivalent time doing part-time work at Walmart!!

But - hacking on computer hardware/software is way more fun....!

 

[Wisecracker]

Elevate the rabbit cage and place your fan underneath -- preferably pointing upward

Fashion a hood over the cage (pretty much anything will do). You could use cardboard, or build a simple frame, and cover it with yer GF's massive panties (it's a joke!). Vent the waste heat to the exterior during hot times, or to an interior register to warm your tootsies during those cold nights.

8-inch insulated flex duct is less than $1US per foot. Ditto for Styrofoam residential insulation sheathing (for the hood). A couple case fans can be used to boost air flow in the duct.

Contractors throw away oodles of this stuff at job sites if you're on a low budget, but please be careful. If you are a hippy-looking type they'll think you're growing Weed, beat you up, and smoke all your Pot.

Or, you could sell all of the GPUs, buy 30 kegs of Molson Ale, throw one heck of a party, and become a legend in your community!

 

[Rob1C]

You definitely want to Flash new GPU Firmware and undervolt to maximize the Hash/Watts and minimize the Heat/Noise - read up on the best Card, BIOS, Settings, etc. for maximum efficiency and either move to Winterpeg or Serbia (cold and low cost Electricity).

 

[firebirdude]

Ether is high as of this post ($293), but still, most regions would only see ~$175-$200 profit after electricity is removed. Also, the majority of us don't have the luxury of a dank cold basement. So this extra cooling would eat into that $175 a bit. NOR do we have multiple decent graphics cards laying around. So if we dump $900 into equipment, it would take nearly 6 months to break even.....even at today's higher Ether rate. Point being, unless Ether starts running at $400+, it ain't worth the headache/stress/time/space used for 99.9% of us.