• Now's your chance win big! Join our community and get entered to win a RTX 2060 GPU, plus more! Join here.

    Pi Cast Episode 3 streams live on Tuesday, August 4th at 2:30 pm ET (7:30 PM BST). Watch live right here!

    Catch Scharon on the Tom's Hardware Show live on Thursday, August 6th at 2:00 pm ET (7:00 PM BST). Click here!

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

Status
Not open for further replies.

shrapnel_indie

Distinguished
Interesting read. Unfortunately for me, I don't have the luxury of hanging to test products (for free, especially)... then again I don't have anything to test, beyond what I grab for my own use. I've got old hardware, but only one GPU with more than a Gig of RAM and I need that for my everyday driver.
 

TJ Hooker

Glorious
Ambassador
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.
 

doktorv

Prominent
Aug 15, 2017
2
0
510
0
If you live in a hot region and use air conditioning, you actually consume more than double the listed electricity because the heat being produced in the computer has to be subsequently removed by the air conditioning, which is not 100% efficient. Texas might seem to be attractive for mining because the electricity is cheap, but if you do the full energy calculation you'll realize it's only worthwhile in the winter.
 

Blake_24

Commendable
Sep 2, 2016
38
0
1,540
3
If you would plan to run this for a longer time, you should try to lower the clock speed on the GPU core to lower power consumption and increase the memory clock speed to increase hash rate. Since Ethereum is not very dependent on the core clock and basically only the memory clock it would make performance better while also increase the profit margin
 

AnimeMania

Honorable
Dec 8, 2014
300
8
10,815
13
Could you write an article about how the cryptocurrencies get the money to pay to their miners. What they do with all the computations your graphic cards are making and if every time they increase the difficulty of mining a coin, does the additional calculations produce something of value or just create busy work. How safe is it allow somebody you don't know to have unlimited, unmonitored access to you computer.
 

TJ Hooker

Glorious
Ambassador

Yes, you need additional power to remove the heat via AC, but it's not double. An AC unit does not take 1 W to remove 1W of heat.
 

takeshi7

Honorable
Nov 15, 2013
105
2
10,685
0
TJ Hooker, an AC unit takes MORE than 1W of electricity to remove 1W of heat. Air conditioning (and all heat exchangers for that matter) are not energy efficient. 2nd law of thermodynamics, and all that jazz. If he's pumping 1000W of heat into the room, he's probably using close to 1500W-2000W of air conditioning to offset it.
 

takeshi7

Honorable
Nov 15, 2013
105
2
10,685
0
When you measured the Hashes/Watt efficiency, did you leave the cards at stock settings? I think you could have gotten much better efficiency results if you tuned the power levels of the cards.
 

TJ Hooker

Glorious
Ambassador

No, they don't, I don't think you fully understand how a heat pump works.

One measure of AC efficiency is EER. It's a measure of cooling capacity (in BTU per hour) by input power (W). A typical AC unit might have a EER of 10. If you convert the units, that's 2.9 W of heat removed per watt of power consumption by the unit. So your cooling costs would result in an extra ~33% on your power bill, not an extra 100% (double).
https://energy.gov/energysaver/room-air-conditioners
 

Ncogneto

Distinguished
Dec 31, 2007
2,349
52
19,860
0
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 your costs?
 

RedJaron

Splendid
Moderator
Because if you only spend $40 or less a month on power, you're still netting $10 / month.
 

extremepcs1

Prominent
Aug 15, 2017
10
0
510
0
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.
 

TJ Hooker

Glorious
Ambassador

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.

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

ubercake

Splendid
Moderator
Does the mining process use much data bandwidth over an internet connection? If so, this would have to be added to the calculations in regions such as the outskirts of the Metro Detroit area where cable providers have no competition and have begun to implement data caps on cable internet.
 

extremepcs1

Prominent
Aug 15, 2017
10
0
510
0
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.
 

doktorv

Prominent
Aug 15, 2017
2
0
510
0
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.
 

shrapnel_indie

Distinguished


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

hons

Distinguished
Jan 8, 2010
16
0
18,510
0
Is there any way to have a "Sample List" of hardware with minimum and maximum hardware requirement ?? Like what combination of CPU, GPU, Memory, ... because I read some articles setup the system with the components they have but I do want to see a "Sample List" which we can add on base on the list. For example, I have a server board w/ 2 CPUs, 2 video cards and 64GB of DDR3 ECC but I don't know if I should put money on the Video cards or add some more memory???
 

none12345

Honorable
Apr 27, 2013
431
2
10,785
0
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)
 
Status
Not open for further replies.

ASK THE COMMUNITY

TRENDING THREADS