FPGA mining for fun and profit

[MoonShadow]

based on the huge difficulty jump today.... I think someone has succeeded in fpga/asic bitcoin flooding

*bell ring*
We have a winner!

Two FPGA clusters went live.

Let me toss out some rough numbers. Please correct me if I'm too far off.

You have about 300 chips doing 200 MHash/s? That's 60 GHash/s you brought online? While that probably puts you in as one of the top miners, that's not a significant cause for the recent network growth.

For someone without piles of existing fpgas sitting around idle: If purchased new, that's $250k in hardware costs bringing in $10k per week in bitcoin value. Half a year to pay off the initial investment (assuming price/difficulty growth remains steady). GPUs would break even in less than half that time.

Don't forget, that was two clusters out of an unknown larger number of them, for which the hardware was already paid for capital.  Those fpga clusters that he speaks of were paid for by other projects, and would otherwise sit idle earning nothing between major projects if not for Bitcoin.  And they are way more energy efficient at hashing than even a GPU, which is way more energy efficient at hashing than a CPU.

[kjj]

I should remind everyone that there has been absolutely zero evidence of any big new arrays, FPGA or otherwise, coming online recently.

Extraordinary claims require extraordinary proof.  So far, there has been zero proof, even of the ordinary kind.

[commlinx]

I should remind everyone that there has been absolutely zero evidence of any big new arrays, FPGA or otherwise, coming online recently.

I was wondering about this recently when I saw a post about a recent dramatic clime in the overall network hash rate...

Am I correct in saying that the hashrate for the network is an estimate based on how many blocks are solved versus the current difficulty? So that while it's statistically likely to be fairly steady there can be lucky and unlucky days even for the whole network combined?

Just editing to add that of course it can also be a lot of new GPU miners coming on-line in combination with the above when only looking over a short period.

[wareen]

Those fpga clusters that he speaks of were paid for by other projects, and would otherwise sit idle earning nothing between major projects if not for Bitcoin.

I would imagine that whoever owns such hardware has some real use for it which he expects to profit from, despite the large initial costs. Bitcoin-mining with these clusters is profitable right now, but since such hardware power can easily be used (and sold) for other profitable projects in the industry, the threshold for them is most probably not electricity cost, but the next most profitable project the hardware could be used for instead.

Don't forget, that with all the home-brew GPU mining rigs only electricity cost is the limit because their computing power can not easily be sold for anything else. If difficulty jumps up, the profitability shrinks for everyone - even those with the most powerful hardware.

Another factor is, that we have a global competition for energy prices and I am pretty sure someone here has access to cheaper electricity than whoever operates those clusters.

Last but not least, GPUs are getting faster and more efficient as well. Due to the much larger market and the very low hardware costs compared to FPGAs or ASICs, the advantage of the latter two shrinks continuously and the cost to upgrade them to the next better hardware generation is vastly greater.

FPGAs and ASICs are by far not the most imminent danger for Bitcoin so I think we should all just relax a bit, congratulate cypherf0x for his current advantage and upgrade our rigs with Southern Islands when they are available.

Everything's going to be just fine 

[Zerbie]

The price of the cards I use for GPU mining initially cost me $270 per.  Comparable cards cannot be found on newegg.com, and the "out of stock" price listed is >$500 per card.  AMD and NVIDA cards are just not competing.  My bet is that ATI knows exactly why their cards are selling so well and would be willing to create low cost ASIC, FPGA, or GPU hardware specifically targeted to mining.  If I were NVIDA or AMD, I would be searching for ways to cater to this market as well.  

If you are really concerned about a threat to FPGA centralization of computing power, start an email campaign to NVIDA and AMD asking them to create hardware specific cards to meet our mining needs.  All that is needed is a bit more competition.

EDIT: Cleaned up awkward wording.

[Transisto]

My bet is that ATI knows exactly why their cards are selling so well and would be willing to create low cost ASIC, FPGA, or GPU hardware specifically targeted to mining.  If I were NVIDA or AMD, I would be searching for ways to cater to this market as well.  

Right now that's no more than 10 000 AMD GPU mining,   Does that really matter for AMD ?

[ttul]

The price of the cards I use for GPU mining initially cost me $270 per.  Comparable cards cannot be found on newegg.com, and the "out of stock" price listed is >$500 per card.  AMD and NVIDA cards are just not competing.  My bet is that ATI knows exactly why their cards are selling so well and would be willing to create low cost ASIC, FPGA, or GPU hardware specifically targeted to mining.  If I were NVIDA or AMD, I would be searching for ways to cater to this market as well.  

If you are really concerned about a threat to FPGA centralization of computing power, start an email campaign to NVIDA and AMD asking them to create hardware specific cards to meet our mining needs.  All that is needed is a bit more competition.

EDIT: Cleaned up awkward wording.

I'm sure some product line manager at AMD has noticed the strange selling out of 6990s and 5970s, and if they have even an ounce of grey matter they've deduced it's the result of Bitcoin. But to support the development of a custom ASIC by one of the big boys, the card sales volume needs to be truly astronomical. I don't see Bitcoin really changing the course of NVIDIA or AMD until there can be provable demand for millions of such custom cards.

In the mean time, there is definitely a niche market for a custom Bitcoin ASIC, if someone can get the job done profitably.

[Steve]

What are the odds of getting the next block, and being able to prevent transactions that you don't approve of for ~10 minutes?  51%

You don't get it. If you have more than 50% of the mining power you can just continue searching for a valid block even after somebody else has found one before you. This way you can make sure that you control every block. This is listed in the weaknesses on the Wiki. Yes, you can technically still add transactions, but there's not much point if they'll never get confirmed.

No, he gets it, and you are still a little behind on the curve.  Having control of 51% of the hashing power of the whole network makes it possible to successfully attack the blockchain for a short period of time.  That period of time being one 10 minute interval.  The whitepaper doesn't go into detail about the odds of success of such an attack, other than to show how it's not really possible at all at less than 50%.  Having just over 50% of the network hashing power doesn't really give you very good odds of success past one block, and an attacker intending to deny transactions into the blockchain for longer than one block has to be able to be certain that no blocks can sneak in under him, for if one gets in and the next is built on top of that before he build one to overwrite that one and one on top to secure his false one then it become exponentially more difficult for him to overwrite two blocks back.  In practice, an attacker wishing to keep this up for an extended period of time needs at least double the hashing power of the network because it's like the attacker is trying to wade up river while the honest nodes are wading down river.  And even with double the rest of the network, some blocks are going to slip in and be covered up again anyway.  At which point the attacker has to choose between trying to overwrite two blocks and then write another before a third is made by the network or simply ignore the one that got away and overwrite the last one to take the network back.

This is not correct.  You have to remember that the chain with the highest difficulty is the one that the network accepts.  If you control >50% of the mining power, you do not have to build on anyone else blocks, you only have to build on top of the ones you produce.  Over time, your chain pull away from the chain that the rest of the network is producing and your chain will be *the* bitcoin chain.  This attack would afford you the opportunity to double spend, or block transactions that spend out of certain accounts, or block transactions altogether...and probably many other kinds of mischief.

[anisoptera]

What you are missing about this attack is that you have to start your chain the minute you give someone coins. You can't precompute a bunch of blocks because to double spend, you have to first spend.

So you send someone money and now you're working on a chain. If they wait a few confirms (the paper describes in detail how quickly the efficacy of this attack drops off with number of confirms) then you have to have beaten the network for X confirms. That is, your hidden blockchain fork has to be at least X+1 blocks long when they release whatever they are releasing to you.

The chance of that happening increases as your power goes up in relation to the network, it's true. But you are working on a blockchain completely independently of the network. You want to do a lot more than equal ("51%") the network - you want to at least have 1.5x the power. You need to be able to find blocks faster than the entire rest of the network, and you need to beat variance and do it consistently for long enough for your transaction to be confirmed "enough" for whatever you're doing.

And if the honest network happens to get luckier than you and get ahead early, you're screwed. Catching up gets exponentially less likely with every block they're ahead. If they get two blocks ahead, well, now you've basically wasted a bunch of processing power finding all those blocks that you COULD have used to generate blocks instead. Maybe your scam would have paid much much more than that, but if you fail you're out the bitcoins (i.e. the extremely large amount of bitcoins you're trying not to spend, getting spent) and the generation profits.

[JJG]

I'm sure some product line manager at AMD has noticed the strange selling out of 6990s and 5970s, and if they have even an ounce of grey matter they've deduced it's the result of Bitcoin. But to support the development of a custom ASIC by one of the big boys, the card sales volume needs to be truly astronomical. I don't see Bitcoin really changing the course of NVIDIA or AMD until there can be provable demand for millions of such custom cards.

In the mean time, there is definitely a niche market for a custom Bitcoin ASIC, if someone can get the job done profitably.

I think you vastly underestimate the number of non-bitcoin users of these cards. At best, bitcoin would be lumped in with the other distributed computing projects that can use GPUs. And don't forget the armies of actual gamers who purchase these cards. Bitcoin, although large, is very much just a drop in the bucket.

Also, you will not see a bitcoin ASIC. Complex ASICs required highly skilled (read: expensive) engineers and even more expensive processes to put them into place. When there is already a very good solution to the problem (GPUs) and an enormously volatile and unpredictable exchange rate, you're not going to see such an investment. Not to mention that investors in such a project would almost certainly demand immediate cashing out of mining proceeds to pay off their investment, which would put significant downward pressure on the exchange rate. ASIC mining is a pipe dream.

[ttul]

Also, you will not see a bitcoin ASIC. Complex ASICs required highly skilled (read: expensive) engineers and even more expensive processes to put them into place. When there is already a very good solution to the problem (GPUs) and an enormously volatile and unpredictable exchange rate, you're not going to see such an investment. Not to mention that investors in such a project would almost certainly demand immediate cashing out of mining proceeds to pay off their investment, which would put significant downward pressure on the exchange rate. ASIC mining is a pipe dream.

I hate to disappoint you...

[Freakin]

Also, you will not see a bitcoin ASIC. Complex ASICs required highly skilled (read: expensive) engineers and even more expensive processes to put them into place. When there is already a very good solution to the problem (GPUs) and an enormously volatile and unpredictable exchange rate, you're not going to see such an investment. Not to mention that investors in such a project would almost certainly demand immediate cashing out of mining proceeds to pay off their investment, which would put significant downward pressure on the exchange rate. ASIC mining is a pipe dream.

I hate to disappoint you...

Unless you're referring to those had to read your vague half sentence, I don't think you've disappointed anyone.

Are you suggesting that you know of someone putting millions of dollars into developing a custom ASIC for mining?

[kjj]

Also, you will not see a bitcoin ASIC. Complex ASICs required highly skilled (read: expensive) engineers and even more expensive processes to put them into place. When there is already a very good solution to the problem (GPUs) and an enormously volatile and unpredictable exchange rate, you're not going to see such an investment. Not to mention that investors in such a project would almost certainly demand immediate cashing out of mining proceeds to pay off their investment, which would put significant downward pressure on the exchange rate. ASIC mining is a pipe dream.

I hate to disappoint you...

Unless you're referring to those had to read your vague half sentence, I don't think you've disappointed anyone.

Are you suggesting that you know of someone putting millions of dollars into developing a custom ASIC for mining?

You mean other than ArtForz?

[molecular]

Also, you will not see a bitcoin ASIC. Complex ASICs required highly skilled (read: expensive) engineers and even more expensive processes to put them into place.

ArtForz designed and ordered himself some ASICs (het put two rounds of sha256 unrolled plus H==0 logic on them, if I got that right, and I don't think one would call 2xsha456 "complex") around January and already received the first 2 batches (100+100) (this might even be old info by now, I think he might've been planning for 1000, not sure).

Some weeks ago the first 96 where already online. They each do 200MHash/s using only 8W. He controls, I think, 8 of them on a small board with an fpga to feed the work and receive the solutions, which he plugs into a backplane of some sort. All 96 fit in a 4U casing.

So your assertion "you will not see Bitcoin ASIC" is wrong, it's already here, ask him on #bitcoin-dev.

And yes: the guy is highly skilled. But "expensive process to put them in place", I don't know: he said he used a modified toaster to do contact soldering

[allinvain]

Question is how much did he pay for the whole process? Tens of thousands of dollars?

200 Mhash per chip with just 8 W is amazing though!

[MoonShadow]

Question is how much did he pay for the whole process? Tens of thousands of dollars?

200 Mhash per chip with just 8 W is amazing though!

IIRC, ARtForz mentioned at one point that his cluster had crossed the $200,000 mark back in the Fall of 2010.  So we aren't talking about your average geek throwing this together in his garage.  ArtForz has resources.

[allinvain]

Question is how much did he pay for the whole process? Tens of thousands of dollars?

200 Mhash per chip with just 8 W is amazing though!

IIRC, ARtForz mentioned at one point that his cluster had crossed the $200,000 mark back in the Fall of 2010.  So we aren't talking about your average geek throwing this together in his garage.  ArtForz has resources.

God daaaamn! 

Yeah, that's one serious bitcoin investment there. It makes my several thousand in mining hardware seem like peanuts. Well, good for him! I do hope he uses his bitcoin wealth to further the bitcoin economy and to start new and interesting ventures. I'd also hope he'd sell me one of his magical ASICS but hmm doubt that will happen

P.S ArtForz, I can pay you in bitcoins.

[Dobrodav]

Bitcoins have failsave  to resist case of "over 50% hash power"  failed in one hands.  I am not remember when, but think satoshi mentioned that in some update notes.

[marcus_of_augustus]

ArtForz is to mining as Satoshi Nakamoto is to software.

The guy is a legend, does he really exist?

Next thing you know there'll be a rumour that he whipped a fab. room in his backyard from old tractor parts and TV sets to pump out ASICs!

[allinvain]

ArtForz is to mining as Satoshi Nakamoto is to software.

The guy is a legend, does he really exist?

Next thing you know there'll be a rumour that he whipped a fab. room in his backyard from old tractor parts and TV sets to pump out ASICs!

Hahahaha... I'd pay to see his mining operation in action. Geek heaven that's for sure!