Profitability of Mining + Excel Workbook

[Cheeseman]

Profitability of BTC Mining

Summary
After stumbling onto Bitcoins about three weeks ago, I immediately started to run numbers to figure out the kind of return I can expect on any kind of hardware investment. I will be looking at two scenarios: upgrading a current desktop computer with an ATI graphics card, and building a new computer for the sole purpose of BTC generation. I will be using my own experimental data, along with data found on the wiki for my calculations. I used Slush's pool for two weeks to determine averages, and all numbers have been updated for the current difficulty. There is no real difference between solo, Slush, or Deepbit in the long run, so this may be irrelevant.

Profitability of Upgrading
I decided to test only the profitability of the 5870 and 5970, as those are the most popular high end GPUs. Not only that, but the cost to expected ROI is the greatest for these two cards. These numbers are real experimental numbers, but your results may vary slightly by +-5%. I'm assuming a USD/BTC exchange rate of 1 for now, as that seems to be where the market is heading.

https://i.imgur.com/T0veD.png

As you can see, after around a month and a half, you have paid off your investment in equipment, regardless of what you buy. After that it is pure profit, disregarding electricity costs. The likelihood of bitcoins crashing in the next 45 days is pretty low, so at the least it is a solid investment into a free graphics card (or whatever you spend on one if you sell it). The likelihood of this information being relevant a year from now is very low, so I would take the yearly profit numbers with a grain of salt.

Profitability of Building New Rig
After pricing out about ~10 or so dedicated mining rigs, I have found that the only rig worth the money is the dual 5970 setup. Think about it like this: a mining rig is comprised of two types of components, namely BTC generating and non-BTC generating components. The lowest cost that I could get the non-BTC generating components to be (CPU, case, PSU, mobo, RAM, and HDD) was around $350. With the next best combination of cards being a dual 5870 config, then the ratio of BTC-generating to non-BTC generating is 400/350 = 1.14. With a dual 5970, that comes out to 800/400, which is 2. Therefore the dual 5970 dedicated mining rig setup is the most efficient.

The mining rig:

Code:

$120 mobo (http://www.newegg.com/Product/Product.aspx?Item=N82E16813131402)
$140 psu (http://www.newegg.com/Product/Product.aspx?Item=N82E16817139009)
$50 RAM (http://www.newegg.com/Product/Product.aspx?Item=N82E16820231277)
$40 HDD (http://www.newegg.com/Product/Product.aspx?Item=N82E16822148395)
$55 Case (http://www.newegg.com/Product/Product.aspx?Item=N82E16811119068)
$40 CPU (http://www.newegg.com/Product/Product.aspx?Item=N82E16819103698)
$800 5970x2 (ebay)
$1245 per 1,210,000 khash
~$1 per mhash

Well, how did it do? About the same as the upgrade dual 5970, except with a price tag of $1245 total and a break-even time of 66 days. With this mining rig, I'd like to introduce a new date, called the "bail-out time". The bail-out time, or the time it takes to pay for the non-resell-able parts of the rig, is 22 days. After that you can recover your investment safely.

HPC GPGPU Profitability
So if this is so profitable, then why not invest tens of thousands of dollars into high-performance computing (HPC) for a quick turnaround? There are two problems with this. The first is that PCIE expansion chassis's are extremely expensive, let alone the servers they connect back to. Plus, with a GPU core limit of 4 for Windows and 8 for Linux, you would need more than one server to run a reasonably sized cluster. Let's go back to our BTC generation ratio to examine this further. A quote on a C410x Expansion chassis was $4000, plus a $6000 C6100 server. That will handle up to 8/16 GPUs. Therefore we need three servers, plus the chassis, to make it completely efficient. It costs a total of $6400 for 16 5970s. Therefore 6400/16,000=.4. Not efficient at all. Might as well spend $9600 for eight individual mining rigs to get the most out of your investment.

There ya' go! Here is an Excel workbook whereby you can plug in your own numbers to find the break even time, etc:
http://www.megaupload.com/?d=PVQ4UDZW

Please feel free to dispute my findings, offer new insights, or donate! 17xBRKsw6S2JneSPdckX8dcnQX67TM4V52

-Cheeseman

EDIT1: Workbook updated to take electricity into account. Effects are negligible unless you live in Hawaii.
EDIT2: Updated GPU limit on Linux

[1714191715]

1714191715

[1714191715]

1714191715

[1714191715]

1714191715

[EPiSKiNG]

Sweet spreadsheet.  Can't wait to plug in my numbers to see what happens!

Thanks!

-EP

[Grinder]

This is way too optimistic. Difficulty has been rising much faster than the price, and you need to include electricity.

[Cheeseman]

This is way too optimistic. Difficulty has been rising much faster than the price, and you need to include electricity.

This is why the yearly figures are listed as completely unreliable. However, up until the break even period (1.5 months), the difficulty changes should have small impact on these figures. The workbook should just be used as an estimate, and may be off by as much as 20% depending on the changes in difficulty.

Also, I can factor in electricity costs pretty easily. I had no need to, since I just use my University's free power

EDIT: Workbook updated for electricity costs.

[grue]

you should use cell references instead of hard-coding them into the cell.

[Cheeseman]

you should use cell references instead of hard-coding them into the cell.

Hm? Everything except the basic inputs reference other cells.

[JWU42]

Have the same MB - you shoulda bought it 2 months ago - $98.99 with $20 MIR = $78.99 

[Raulo]

You are overly optimistic and you are rounding all costs down and all benefits up.

Ebay 5970 prices are rather $450 than $400. You would need a better case (or build one yourself). This one will not give you enough airflow. You need way more ventilation and a few good fans ($15-20 each). If not, you will not be able to overclock the cards, at least not to all way up. The system will not need 294x2 W but 700-750W depending on the PSU efficiency. Depending on your climate, you will need to air condition this 700W. The BTC bid price is, last time I checked, $0.87, not $1.00  and nobody knows, where it is headed (it was $0.6x just a few days ago). And if it is indeed $1.00 or more, you can be certain of a significant difficulty increase. Before you even order these parts, the difficulty is going to be 10-15% more and when you have it all running, you will be half way to another 10% of difficulty increase. Taking all that into account (especially the difficulty increase of 10% every 2016 blocks which may be conservative but during all my time with Bitcoin, difficulty increased faster than I always expected), you are going to get 1500 BTC before electricity costs in 180 days (minus 3 BTC a day of electricity costs with AC=540 BTC) which is not going to pay for this system at current BTC prices. Then, another 90 days will bring you only 300 BTC before electricity (but maybe the difficulty increase will slow and it will be a bit more). Did I mention that your eBay card may fail after a few weeks of constant use and you are out of warranty?

In my opinion, it is well worth mining with a system you already have. And it is also worth building a mining rig if you like building computer systems and monitoring it because it's fun. And the mining will eventually pay for the system (but not for the time you spend on it so consider it a hobby). But profit? Not so much unless the BTC price increase well beyond 1$ (and better yet not now but later so it does not impact the current difficulty). And for the prices, we are in the "tea leaves reading" territory. And if you expect higher prices, it is much simpler to just buy BTC, instead of the mining rig.

[Zamicol]

Funny, you sound just like me!  I completed a spreadsheet over he weekend and I just finished ordering some parts for a new machine, but It's going to be used for work/play too, so I spent a little more.  My number are a little screwy, I made them to factor in parts that I already had depending on hardware demands.  I'm shooting for dual or triple 5870's. 

When I did my calculations, my payback was more conservatively at 90 days.  This is anticipating a very small increase in difficulty and not over clocking the cards as you were.  I didn't factor in electricity. 

I have a few comments as well.
That case is ugly and it cost ~$20 for shipping.  May I suggest http://www.newegg.com/Product/Product.aspx?Item=N82E16811147023? It's 10 bucks less, looks like it might have better airflow, and free shipping!

And 5970's at even $450 I havn't seen for a while on Ebay.  It's all you darn mining folk.

And by the way, 5870's @ 370 Mhash/s?  Did you kill something to achieve this?

I seriously almost bought that Mobo.  I ended up buying this:http://www.newegg.com/Product/Product.aspx?Item=N82E16813131647&cm_re=ASUS_M4A87TD_EVO-_-13-131-647-_-Product

I bought a little nicer case, it was on sale and I have wated to get a decent one for a few years now.  A great excuse  http://www.newegg.com/Product/Product.aspx?Item=N82E16811119196

Good luck. 

[Grinder]

This is why the yearly figures are listed as completely unreliable. However, up until the break even period (1.5 months), the difficulty changes should have small impact on these figures. The workbook should just be used as an estimate, and may be off by as much as 20% depending on the changes in difficulty.

You include an expected rise in value but not in difficulty, which is typical cognitive bias. Although the price has risen to about .95 last night (still less than your expected price), the short term mining power has had very sharp rise as well. At the current level, it would result in a 75% increase in difficulty. My guess is that it won't be that bad for a while, but then the price may go down again as well.

[eche|on]

Hi!

just one note:
over here I pay 30 cent per kw/h power. A 300 MHashes/sec GPU consumes 300W. 3hours mining is roughly 30 cent.
8 times a day are 24h are roughly 2€ per day power costs.
A 300 MHashes/sec at current diff (83.000, will be 92.000 next week) will give me 3.6 BTC per day. Even at a rate of 1 US$ per BTC thats only 3.6 US$ or 2.5€ per day. Which is roughly 50 €-cent per day on plus side.
30 days a month are 15 € per month. For a 300 MH card I pay 330 €, which will be 22 month of mining with current difficulty until card is payed (and power).
Somehow something does not fit here. And difficulty does raise.
So, how will you gain 6000$ per year with investment cost, raising difficulty AND power costs?

eche|on

[mrb]

Plus, with a GPU core limit of 4 for Windows and 6 for Linux.

On Linux, you can use up to 8 AMD GPUs, not 6.
http://blog.zorinaq.com/?e=42

[rezin777]

And by the way, 5870's @ 370 Mhash/s?  Did you kill something to achieve this?

This doesn't require voodoo.

[rezin777]

over here I pay 30 cent per kw/h power. A 300 MHashes/sec GPU consumes 300W.

That's very expensive electricity. And there are cards that can do 350MHash/s at 180W.

[Littleshop]

And by the way, 5870's @ 370 Mhash/s?  Did you kill something to achieve this?

This doesn't require voodoo.

What does it require?  I get 330 with 900mhz clocks, i know i could go higher on the clock a bit but choose not too. 

[deadlizard]

And by the way, 5870's @ 370 Mhash/s?  Did you kill something to achieve this?

This doesn't require voodoo.

My first thought

 

[rezin777]

And by the way, 5870's @ 370 Mhash/s?  Did you kill something to achieve this?

This doesn't require voodoo.

My first thought

 

This is before my time (computer wise), but awesome none-the-less!

[Cheeseman]

You are overly optimistic and you are rounding all costs down and all benefits up.

Ebay 5970 prices are rather $450 than $400. You would need a better case (or build one yourself). This one will not give you enough airflow. You need way more ventilation and a few good fans ($15-20 each). If not, you will not be able to overclock the cards, at least not to all way up. The system will not need 294x2 W but 700-750W depending on the PSU efficiency. Depending on your climate, you will need to air condition this 700W. The BTC bid price is, last time I checked, $0.87, not $1.00  and nobody knows, where it is headed (it was $0.6x just a few days ago). And if it is indeed $1.00 or more, you can be certain of a significant difficulty increase. Before you even order these parts, the difficulty is going to be 10-15% more and when you have it all running, you will be half way to another 10% of difficulty increase. Taking all that into account (especially the difficulty increase of 10% every 2016 blocks which may be conservative but during all my time with Bitcoin, difficulty increased faster than I always expected), you are going to get 1500 BTC before electricity costs in 180 days (minus 3 BTC a day of electricity costs with AC=540 BTC) which is not going to pay for this system at current BTC prices. Then, another 90 days will bring you only 300 BTC before electricity (but maybe the difficulty increase will slow and it will be a bit more). Did I mention that your eBay card may fail after a few weeks of constant use and you are out of warranty?

In my opinion, it is well worth mining with a system you already have. And it is also worth building a mining rig if you like building computer systems and monitoring it because it's fun. And the mining will eventually pay for the system (but not for the time you spend on it so consider it a hobby). But profit? Not so much unless the BTC price increase well beyond 1$ (and better yet not now but later so it does not impact the current difficulty). And for the prices, we are in the "tea leaves reading" territory. And if you expect higher prices, it is much simpler to just buy BTC, instead of the mining rig.

The dedicated mining rig is the riskiest option out of the two that I posed. The fact that this market can gain/lose 40% of it's value in the course of a week is insane, and makes it a risky investment. I absolutely recommend just upgrading, as you have less time (and therefore less risk) until you hit the break even point. And even then, you can "get out" at any time by just selling the card itself.

Trust me, I don't have the capital nor the balls to build a dedicated mining rig. I have upgraded two computers and nearly have them paid off (sold off all of my BTC in the last day). This just shows that it is possible to make a short term profit mining bitcoins. Is it a risky investment? Hell yeah, as there are a number of things that could go wrong. Is it safer than speculating in BTC itself? In my opinion, absolutely. I would rather put $1000 into a mining rig and have the fallback option of selling it if the market tanks, rather than owning a lot of worthless currency.

And to all of the people attacking me on difficulty: can you recommend a way to "guess" at a better break even point using a difficulty estimate? I'm not concerned with anything past the break even point (~45 days), as that is when it becomes purely profitable to run your rig regardless.

And also, yes, I get 370,000 khash on my 5870 on air through overclocking. The average between my two currently is about 360000 khash, and the hottest temp I have is 70 degrees celcius.

[rezin777]

And by the way, 5870's @ 370 Mhash/s?  Did you kill something to achieve this?

This doesn't require voodoo.

What does it require?  I get 330 with 900mhz clocks, i know i could go higher on the clock a bit but choose not too. 

It requires going higher on the clocks!   

Playing with other settings. Finding the right combo of drivers / sdk. Reading the forum and see what works for others. Basically try different things until you see an improvement.

[Littleshop]

And by the way, 5870's @ 370 Mhash/s?  Did you kill something to achieve this?

This doesn't require voodoo.

What does it require?  I get 330 with 900mhz clocks, i know i could go higher on the clock a bit but choose not too. 

It requires going higher on the clocks!   

Playing with other settings. Finding the right combo of drivers / sdk. Reading the forum and see what works for others. Basically try different things until you see an improvement.

Ok.  900mhz and 330M is fine with me then.  I do not want to push my cards over 900mhz.