Mining profitability poll

[Ian Maxwell]

I'm getting wildly different estimates from one day to the next, with regard to how long mining will be profitable. I'm not planning to build a high-powered rig---just to keep running my one 5970 until I start losing money---but I don't think the revenue/cost ratio is much higher on two cards than one, so that shouldn't affect the answer.

If it's long enough, maybe I can spend some of my profits on another card!

[Jaime Frontero]

There's only one variable that has any meaning.  Exchange rate:  short-term, mid-term, and long-term.

Everybody has a different answer for that.  And nobody's got a lock on what it's really going to be.

I see no difference at all between Bitcoin miners, and the old '49ers at their gold sluices.  Some may be more cynical than others - but we're all optimists, betting on the come.

So what, if your mining costs are $2.00/BTC - when you're planning on holding for two years and you think the exchange rate will go up to $10.00?  Under that assumption, you're laughing all the way to the bank.

[JJG]

...
So what, if your mining costs are $2.00/BTC - when you're planning on holding for two years and you think the exchange rate will go up to $10.00?  Under that assumption, you're laughing all the way to the bank.

You can BUY BTC for $1.60 / BTC. Mining at $2.00/BTC would be less profitable than buying them outright. In this case, the hardware vendors and electricity companies are laughing all the way to the bank.

If you really think the exchange rate will continue upward forever, then joining on the mining bandwagon for the race to the bottom is not a good business decision.

[proudhon]

I'm getting wildly different estimates from one day to the next, with regard to how long mining will be profitable. I'm not planning to build a high-powered rig---just to keep running my one 5970 until I start losing money---but I don't think the revenue/cost ratio is much higher on two cards than one, so that shouldn't affect the answer.

If it's long enough, maybe I can spend some of my profits on another card!

To project profitability you need to know what your inputs are. Once you know these costs you also have to be able to guess about what  $/BTC and Difficulty will be in the future.

I recommend starting a spreadsheet and coming up with some projections that you can believe based on what Bitcoin has done so far. I have done this and continue revising my presumptions as I learn more about how Bitcoin and its markets work. To start, I would plot some basic information like block number, price, and difficulty at least as frequent as every re-targeting period. Then you can use this info to see what it looks like using various moving averages and trends. Use what you know to be true about the mathematics of Bitcoin and build on that to come up with a business plan before you spend too much on it.

This really helps with drowning out the noise of daily fluctuations in price and sentiment.

I think this is good advice.  For the past few days I've been making this sort of spreadsheet.  The big questions I want it to answer, which is currently does answer are:

1.  What is my current profit at the current exchange rate per hour, day, month, year?  Given certain values for Mhash/s, reward/block, difficulty, cost/kWh, and USD/BTC, my spreadsheet calculates this for solo mining and PPS mining with Deepbit (essentially, the average minus 10%).  The month and year projections aren't very useful because the relevant variables (e.g. difficulty, exchange rate, etc.) don't stay constant enough.  Difficulty, for example, is guaranteed to change at least every 2 weeks.

2.  What will the exchange rate of USD/BTC need to be at the next projected difficulty level in order to retain the same profit as at the current difficulty level and exchange rate?  For example, if the difficulty goes up by 15.38% (i.e. 106,552), which is the current projection at bitcoincharts as of this posting, then the exchange rate would need to go up to ~$1.92 in order for me to make the same amount I'm making at the current difficulty level with an exchange rate of $1.66.

3.  What would my profit be if the difficulty increases, but the exchange rate does not?

4.  Finally, at a given exchange rate, at what difficulty would I only be making my hourly power cost, and, thus, zero profit?  For example, with my hardware and power costs, if the exchange rate remained the same, then at difficulty 617,779 I would make enough to merely cover the cost of the electricity used to mine.  I think this is about as close as you can get to predicting how long it will take before a given hardware configuration is unprofitable.

There are lots of other questions I want it to answer and I'm constantly adding new formulas, but these are the big ones right now.  Of course, it tells me lots of other stuff like my BTC/hour, BTC/hour minus PPS fee with deepbit, hours per BTC with and with PPS, gross USD/hour, days and months to cover the cost of hardware, power costs per hour, day, week, month, and year, and so on.

[fpgaminer]

Is there a place to get the difficulty history? The site linked from the wiki, http://nullvoid.org/bitcoin/difficultiez.php, appears to be down at the moment.

Also, specifically for the poll, do you mean to include the capital cost (cost of buying the cards)? For example, if you already own your cards and don't care about getting your money back on them, then they will be profitable as long as they generate more than the maintenance costs. That will likely persist for quite a bit longer than the same figures after factoring in capital costs.

[JJG]

Is there a place to get the difficulty history? The site linked from the wiki, http://nullvoid.org/bitcoin/difficultiez.php, appears to be down at the moment.

Also, specifically for the poll, do you mean to include the capital cost (cost of buying the cards)? For example, if you already own your cards and don't care about getting your money back on them, then they will be profitable as long as they generate more than the maintenance costs. That will likely persist for quite a bit longer than the same figures after factoring in capital costs.

http://bitcoin.sipa.be/ has the best graphs, but the server is incredibly flaky. If it doesn't load, try again in 10 minutes.

Don't forget that these cards were not designed to be run full-tilt 24/7. Loading it to 95+% 24 hours per day will absolutely make it fail early. Still, the early failure could be so far out mining will be unprofitable (electricity costs only) by then.

[Jaime Frontero]

I'd say there are 4 variables not 1. Namely:

- exchange rate
- difficulty
- capital costs i.e. equipment etc..
- recurring costs i.e. electrcity, rent, wages etc...

You're right, vladimir, of course.

Perhaps I should have said "the non-calculable variable is the only one that matters..."

[mjsbuddha]

[fpgaminer]

Just because I'm a whore for my little project: My puny little FPGA miners will remain profitable until the difficulty reaches ~4,000,000, assuming 1BTC = $1USD and $.15 per kWh. Go ahead, check my math:

80MHash@5Watts
5 W * 24hr => 0.12 kWh/day
0.12 kWh/day * $0.15/kWh => $0.018/day
50 BTC / x days => BTC/day; Need 0.018 BTC/day assuming 1BTC=1USD
50 BTC / 2777 days = ~0.018 BTC/day (profitable)

4000000 Difficulty @ 80MHash/s = Average 2485 days for payout

 

[proudhon]

Is there a place to get the difficulty history? The site linked from the wiki, http://nullvoid.org/bitcoin/difficultiez.php, appears to be down at the moment.

Also, specifically for the poll, do you mean to include the capital cost (cost of buying the cards)? For example, if you already own your cards and don't care about getting your money back on them, then they will be profitable as long as they generate more than the maintenance costs. That will likely persist for quite a bit longer than the same figures after factoring in capital costs.

The best place to get difficulty history is the block chain itself... Use: http://blockexplorer.com/

These are the relevant Blocks, total number of Bitcoin, and Difficulty to the nearest integer:

2016    100,800    1
4032    201,600    1
6048    302,400    1
8064    403,200    1
10080    504,000    1
12096    604,800    1
14112    705,600    1
16128    806,400    1
18144    907,200    1
20160    1,008,000    1
22176    1,108,800    1
24192    1,209,600    1
26208    1,310,400    1
28224    1,411,200    1
30240    1,512,000    1
32256    1,612,800    1
34272    1,713,600    1
36288    1,814,400    1
38304    1,915,200    2
40320    2,016,000    3
42336    2,116,800    4
44352    2,217,600    5
46368    2,318,400    5
48384    2,419,200    6
50400    2,520,000    8
52416    2,620,800    11
54432    2,721,600    13
56448    2,822,400    12
58464    2,923,200    17
60480    3,024,000    17
62496    3,124,800    19
64512    3,225,600    24
66528    3,326,400    45
68544    3,427,200    182
70560    3,528,000    244
72576    3,628,800    352
74592    3,729,600    512
76608    3,830,400    623
78624    3,931,200    713
80640    4,032,000    918
82656    4,132,800    1319
84672    4,233,600    1378
86688    4,334,400    2149
88704    4,435,200    3092
90720    4,536,000    4536
92736    4,636,800    6867
94752    4,737,600    8078
96768    4,838,400    12252
98784    4,939,200    14484
100800    5,040,000    16307
102816    5,140,800    18438
104832    5,241,600    22012
106848    5,342,400    25998
108864    5,443,200    36460
110880    5,544,000    55590
112896    5,644,800    76193
114912    5,745,600    68979
116928    5,846,400    82347
118944    5,947,200    92348

Did GPU mining start in between blocks 66,528 and 68,544?  The difficulty increased by 304%.

[ribuck]

Did GPU mining start in between blocks 66,528 and 68,544?  The difficulty increased by 304%.

That difficulty increase wasn't due to GPU mining, which came in gradually over many months.

That 304% difficulty increase seems to have been caused by a guy called Nenelod who worked at a hosting company. He claimed to be using 1000 CPU cores to generate bitcoins (although the actual number was probably lower).

Here's the forum thread about it:
http://bitcointalk.org/index.php?topic=431.0
and here's Nenolod's blog post about it:
http://nenolod.net/thoughts-on-bitcoin/

[Jaime Frontero]

http://twitter.com/#!/nenolod/status/18459913859

#smoothmove

Wotta maroon (to quote the inestimable Bugs Bunny)... 

[fpgaminer]

These are the relevant Blocks, total number of Bitcoin, and Difficulty to the nearest integer:

Wonderful! Thank you for putting that together and posting it!

I'm new at this, but according to Eureqa, optimizing for Root Mean Squared Error, the best fit is:

Code:

f(x) = (2.55e-5 * x)^10.4

Where x is the block number.

I also re-ran it in terms of the difficulty period (so, blocks / 2016), since that makes more sense:

Code:

f(x) = (0.0525 * x) ^ 10.2

Difficulty period is 14 days.

So that can be used to predict the difficulty in the future. I guess it will be accurate enough until mining begins to lose appeal.

Here its predictions, for the curious:

05/02/11      120995.06
05/16/11      143215.23
05/30/11      169051.93
06/13/11      199020.81
06/27/11      233701.03
07/11/11      273741.81
07/25/11      319869.65
08/08/11      372896.14
08/22/11      433726.35
09/05/11      503368.02
09/19/11      582941.4
10/03/11      673689.92
10/17/11      776991.66
10/31/11      894371.74
11/14/11      1027515.63
11/28/11      1178283.46
12/12/11      1348725.4
12/26/11      1541098.12
01/09/12      1757882.5
01/23/12      2001802.49
02/06/12      2275845.43
02/20/12      2583283.6
03/05/12      2927697.41
03/19/12      3313000.01



Now I need to find some BTC<->USD data and model that too 

[Veldy]

I'd say there are 4 variables not 1. Namely:

- exchange rate
- difficulty
- capital costs i.e. equipment etc..
- recurring costs i.e. electrcity, rent, wages etc...

Also worth keeping in mind is whether heat generated can be used positively or not.  In northern states, for the majority of the year, the heat generated helps lower the heating bill, so you get much more efficiency out of your power usage.

[AmpEater]

Indeed, I removed my furnace and heated this winter with nothing but ~5GHash of miners. 

[Ian Maxwell]

By profitable I meant profitable on the margin: as in, when will it be time for me to shut off my miner. (As others have pointed out, if the exchange rate for BTC drops below your marginal mining expenses, continuing to mine is about as profitable as buying bitcoins above market rate.)

[Enky1974]

If that difficulty forecast is correct, here in italy due to high electricity cost already in july 2011 mining with an ati 5870 at 370mh/sec will be unprofitable (with 1btc=1.8usd). Over 250000 difficulty factor i'll have to stop mining but i hope that at some point difficulty will drop due to unprofitability or viceversa btc will increase in value to compensate,at the moment i dont have any idea if price drives difficulty or viceversa.

[phelix]

Just because I'm a whore for my little project: My puny little FPGA miners will remain profitable until the difficulty reaches ~4,000,000, assuming 1BTC = $1USD and $.15 per kWh. Go ahead, check my math:

80MHash@5Watts
5 W * 24hr => 0.12 kWh/day
0.12 kWh/day * $0.15/kWh => $0.018/day
50 BTC / x days => BTC/day; Need 0.018 BTC/day assuming 1BTC=1USD
50 BTC / 2777 days = ~0.018 BTC/day (profitable)

Your calculation seems correct but I wonder about the 5W. You still need the power for the rig. Or do you have so many FPGAs that it becomes neglectable??

[fpgaminer]

Your calculation seems correct but I wonder about the 5W. You still need the power for the rig. Or do you have so many FPGAs that it becomes neglectable??

No rig is required. The FPGAs can be controlled by a micocontroller (and networking module), and uCs use negligible power. Or the micro can be embedded in the FPGA. You only lose power to the supply, which isn't that bad.

Right now, my computer is acting as the controller. Mostly because I don't have any micros laying around.  

That's nice, but what about capital costs to buy all those FPGA miners?

And there's the rub.

I suspect I can get 160 MH/s out of a Xilinx Spartan-6-150, which is something like $150 in onesies. That might add up to a $300 BOM for a custom board. It doesn't compete directly against a GPU card, but if you factor in that GPUs require an entire rig it might just work. *shrugs* As always, I'll keep people updated on my progress.

And of course a full ASIC spin, or even franken-ASIC (hardcopy) would dominate no problem, assuming available demand to cover the cost of that.

[Veldy]

Indeed, I removed my furnace and heated this winter with nothing but ~5GHash of miners. 

Awesome!