Comparison of GPU's

[Cdecker]

It's been quite some time that I've been wondering whether I dare to tip my toes into GPU mining. But since I only have my working notebook I'd have to buy a dedicated mining rig for it. So just to get an idea about what size of an investment that would be I wanted to ask you what performances can be expected from the various GPU's. To be fair please add the miner you're using on it, as to avoid flaming about which miner is better.

I've heard a value of 625Mhash/s floating around for an ATI Radeon HD5970, is that realistic? Maybe I'll create a comparison matrix from the collected values ^^

So let's get started: what are you getting?

[Essjay]

Previously, someone posted this list of GPU performance:

http://golubev.com/gpuest.htm

A rough estimate of Bitcoin hash speed is: take the figure in "Single SHA1 speed" and divide by four. (How's that?)

[slush]

It's been quite some time that I've been wondering whether I dare to tip my toes into GPU mining. But since I only have my working notebook I'd have to buy a dedicated mining rig for it. So just to get an idea about what size of an investment that would be I wanted to ask you what performances can be expected from the various GPU's. To be fair please add the miner you're using on it, as to avoid flaming about which miner is better.

Don't expect you will get rich. If you want to make money, do something else. I bought 5970 + separate (poor, singlecore) machine before 14 days and from first calculation, my 500$ will be back _maybe_ after 5 months. Completely pointless, until you do not consider that as interesting hobby. Difficulty is going up very quickly (it was also original idea behind my mining service - found as much blocks as possible before difficulty will be higher again :-).

I've heard a value of 625Mhash/s floating around for an ATI Radeon HD5970, is that realistic? Maybe I'll create a comparison matrix from the collected values ^^
So let's get started: what are you getting?

I have stable ~580Mhash/s and it can go little better (was over 620 when I tried that), but I don't want to overclock it so far for common mining. I live in flat and decibels and celsiuses are not my friends.

You already know it, following lines I'm writing for newcomers:

Bitcoin is NOT about mining. Bitcoin is about economy. Consider us as freaky geeks which sometimes do something for fun. Dont boycott Bitcoin just because you think we are all speculators. Personally I would be more rich when I spend my time in my job than playing with bitcoin.

[Cdecker]

Previously, someone posted this list of GPU performance:

http://golubev.com/gpuest.htm

A rough estimate of Bitcoin hash speed is: take the figure in "Single SHA1 speed" and divide by four. (How's that?)

Wow this is pretty extensive coverage. But I don't get why you'd divide by 4

@slush: I want to come out +/- 0 in the end, I love playing with new stuff, so it'd be just for entertainment. Problem is that the overhead of buying a new rig is quite high, so I'm trying to find the right balance between expensive GPUs and time to get the investment back.

[BitLex]

This is an even better list,
nice overview of all decent cards, showing expected Mhashes, Mhases/W, almost anything you need (except for the price).

[Essjay]

Previously, someone posted this list of GPU performance:

http://golubev.com/gpuest.htm

A rough estimate of Bitcoin hash speed is: take the figure in "Single SHA1 speed" and divide by four. (How's that?)

Wow this is pretty extensive coverage. But I don't get why you'd divide by 4

Why divide by 4? Well, Bitcoin requires you to do two rounds of SHA-1, per operation. And... I don't know, I suspect that Bitcoin requires you to hash a larger amount of data? (Or is that why it requires you to do two rounds?)

@slush: I want to come out +/- 0 in the end, I love playing with new stuff, so it'd be just for entertainment. Problem is that the overhead of buying a new rig is quite high, so I'm trying to find the right balance between expensive GPUs and time to get the investment back.

I don't think it's possible to make back your money. I bought a Radeon 6870 a month ago when the difficulty was 3000, to play Team Fortress 2. I hoped that I could make some bitcoins back to recoup the cost. I did, but now the difficulty is 12000 and increasing rapidly (http://nullvoid.org/bitcoin/difficultiez.php). If you're in it to make money, there isn't a future for you.

The best reason to do this is because electronic cash is an interesting concept (as slush says above), and because you wanted the GPU anyway to play Team Fortress 2

[Anonymous]

Live in a cold climate and use bitcoin mining to heat your house.

[Mahkul]

I makes more sense if you don't have to pay for the electricity yourself.

[teknohog]

Why divide by 4? Well, Bitcoin requires you to do two rounds of SHA-1, per operation. And... I don't know, I suspect that Bitcoin requires you to hash a larger amount of data? (Or is that why it requires you to do two rounds?)

Bitcoin uses SHA-256 which takes about twice the time for SHA-1.

As for decibels and Celsiuses, see my other post:

http://bitcointalk.org/index.php?topic=2196.msg31306#msg31306

[Cdecker]

I don't think it's possible to make back your money. I bought a Radeon 6870 a month ago when the difficulty was 3000, to play Team Fortress 2. I hoped that I could make some bitcoins back to recoup the cost. I did, but now the difficulty is 12000 and increasing rapidly (http://nullvoid.org/bitcoin/difficultiez.php). If you're in it to make money, there isn't a future for you.

The best reason to do this is because electronic cash is an interesting concept (as slush says above), and because you wanted the GPU anyway to play Team Fortress 2

The rapidly increasing complexity is a good point, is there an estimate on how fast it'll keep on growing? Once everybody's using their GPUs in the free time the complexity increment will slow down, won't it?

At that point longterm speculations might make more sense, I'll delay my investment until then.

[theymos]

ArtForz is planning on buying $200,000 in custom mining hardware. GPU mining will then become unprofitable for everyone. I don't know when he's going to do this (and it's not sure that he will), but I wouldn't risk buying a GPU for mining.

[Cdecker]

Well, that's a massive investment, are we sure that doesn't tip over the balance of Bitcoin? We do need to have the computation distributed among all peers. It's not that I don't trust ArtForz, he's done very much to improve Bitcoin, and he's trustworthy, but accumulating so much computational power in one place strikes me as insecure. Assume he disappears (casuality, governmental intervention, ...) it would leave the difficulty sky high, so high in fact that for months generation would not go back to it's regular pace, making transaction confirmation extremely slow.

So for the good of the network and to keep other miners interested, please do not invest such large amount all at once. You can stay ahead of others, but don't crush everybody else.

[slush]

accumulating so much computational power in one place strikes me as insecure

+1

[ribuck]

I don't think you're likely to make back your money, if you are paying for electricity, and if you need to buy a suitable GPU card.

Here are my figures:

ATI 5870, cost £279.70
305,000 khash/s = average time to generate a block of 48 hours
48 hours at 338 watts (GPU + CPU) equals 16.224 kilowatt hours per block
16.224 kilowatt hours at £0.14 per kW-h equals £2.27 electricity cost per block
Value of a block (MtGox latest bid, less 1% conversion fee): $11.46 or £7.38
After deducting electricity cost: £7.38 minus £2.27 equals £5.11 "profit" per block

At that rate, the card will need to generate 55 blocks before it pays for itself. At the current difficulty rate, that would take 110 days. Even if the difficulty rises by only 15% at every adjustment, the electricity cost will be about £9.18 per generated block in 110 days, so the card would never be paid for. My hunch is that the difficulty will rise much more than 15% per adjustment on average.

Of course, if the price of bitcoins rises enough, the economics of generation changes completely.

[slush]

Of course, if the price of bitcoins rises enough, the economics of generation changes completely.

Does anybody have an idea what happen when even strong GPUs won't be profitable? I'm affraid of monopoly of few entities with extremely large processing power. Not because they will accumulate bitcoins, but because of possible fragility of network itself. It should remain p2p somehow.

Maybe pools are solution, because pool admin does not own processing power and no one entity can turn to monopoly. Still users own the power, they temporary
lend it to pool and can everytime connect to another entity when they don't like pool politics. Or can start mining back standalone.

[theymos]

Well, that's a massive investment, are we sure that doesn't tip over the balance of Bitcoin?

It will (for a while, at least). ArtForz will have 65% of the network.

With this much CPU power, he could actually prevent all other blocks from making it into the chain. I don't think he will do this, though.

Maybe pools are solution, because pool admin does not own processing power and no one entity can turn to monopoly. Still users own the power, they temporary
lend it to pool and can everytime connect to another entity when they don't like pool politics. Or can start mining back standalone.

Pool mining will still be unprofitable electricity-wise. They could be powered by volunteers, of course.

[kwukduck]

Thanks for the GPU list, very helpfull!

I hope ArtForz won't go thru with this, a lot of people get interrested in the first place because of the mining, ofcourse, thats silly and not the point of bitcoin, but it gets the attention for people to look into it and learn more about it.
And ofcourse to keep the network balanced this wouldn't help.

[slush]

lot of people get interrested in the first place because of the mining

I don't think so. Personally I'm mining alone far around. And I started after few months after I found beauty of bitcoin. Most people I have talked were interested more in economy and stuff around and I'm excited how many people get the basic idea. We definitely need more real merchants and services accepting bitcoins to break our image as hash crunching idiots :-). So I'm looking for pizza4btc stuff a lot .

[Mahkul]

Still, there are not many nicer things to see in the morning on the screen of your PC than:

Generated (matures in XX blocks).

[Essjay]

Of course, if the price of bitcoins rises enough, the economics of generation changes completely.

Does anybody have an idea what happen when even strong GPUs won't be profitable? I'm affraid of monopoly of few entities with extremely large processing power. Not because they will accumulate bitcoins, but because of possible fragility of network itself. It should remain p2p somehow.

If it's not profitable for individuals like us, it won't be profitable for monopolies either. Unless they expect to make their profit in different ways.

If ArtForz wishes to spend $200K of his own money on generation hardware, more power to him (pun not intended). Money talks and he is putting it up.

If it were me, I would consider the risk unacceptable of people leaving the system, and Bitcoin collapsing, and being left with $200K of rapidly depreciating and otherwise useless hardware.