FPGA mining for fun and profit

[SgtSpike]

I still find absolutely no evidence of a massive hashing machine, whether powered by unicorn dust or by FPGAs.

One person has been making very sketchy claims, which of course he can't provide any evidence for because the Men In Black offered him 25% to keep quiet.  That is 25% of a venture that is highly unlikely to be profitable even if they didn't have to pay him hush money, by the way.

If you think that the recent jump in difficulty was caused by a mythical super-cluster coming online in the last day or two, you really need to go back and read up on the retargeting process.

Actually it was 25% and an NDA on the implementation.

If I did a screencast of the system then someone could say it was faked with a script too.  In the end there are always people who don't believe despite whatever you show them.

So you did bring a cluster online?  What sort of GH/s is this cluster putting out?

[1714822037]

1714822037

[1714822037]

1714822037

[1714822037]

1714822037

[1714822037]

1714822037

[sturle]

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.

I can't really describe the feeling of learning that the two initial units were for prototyping and there were RACKS of them to be used when idle.

Are you sure they are on?  I can't see a noticeable bump in total network power.  The increase is steady within normal variance.  Perhaps a lot of GPU miners shut down their systems at the exact same time?  Sure you didn't mix hash/s and Mhash/s?

[ryepdx]

if that kind of mining power gets utilized it will completely destroy the BTC

edit...

sorry the terahash / sec that was quoted earlier...

Wrong. If only one person or entity utilizes that kind of power, that could destroy the bitcoin. What we need is a public service to lower the barrier to entry for people. If FPGA miners are as cost-effective as the Th/s figure quoted earlier, then it would probably be quite profitable for a person to invest in an FPGA mining business. In the long run, it's really just the cost of keeping a cluster running that matters most, assuming you can make back your initial investment.

So you did bring a cluster online?  What sort of GH/s is this cluster putting out?

I second the question.

[edit]

If I did a screencast of the system then someone could say it was faked with a script too.  In the end there are always people who don't believe despite whatever you show them.

True. But could you humor us anyway? Your claims would have a bit more veracity to them if you provided evidence. As they say on teh interwebz: pics, or it didn't happen.

[marcus_of_augustus]

Apologies but no more development information will be posted.  I've been offered a 25% share from someone that owns 2 FPGA clusters.  If you haven't seen that type of hardware before think a 156 FPGAs per machine.

Okay, so the most likely existing FPGA user with this kind of hardware is either a crypto set-up or banking/financial house that used it for algo trading.

So I'm gonna have a guess that it was an existing financial entity of some sort ... which means they will have an incentive to set-up a decent bitcoin exchange with easy interface with existing currency markets to go along with their mining operation. Basically a 'rogue' from the current oligopolies is breaking ranks and going to "have a go". If this is the case, it is great news.

Edit: oy yeah, if there is one finance house out there ready to mobilise their existing FPGAs to spin a bitcoin profit there will be dozens ... a lot of these guys are "on the bones of their arse broke" after the financial collapse of fiat money, just hanging out for every QE injection from the bernank.

[ryepdx]

So I'm gonna have a guess that it was an existing financial entity of some sort ... which means they will have an incentive to set-up a decent bitcoin exchange with easy interface with existing currency markets to go along with their mining operation. Basically a 'rogue' form the current oligopolies is breaking ranks and going to "have a go". If this is the case, it is great news.

Okay, yes. But it's very, very hard good news to handle for those of us who just dropped a couple grand on mining equipment only a few days ago. I want to at least make back my investment before FPGA mining makes it all worthless! 

[marcus_of_augustus]

I warned earlier ... can't find the post, about getting too heavily into GPUs because of threat they would go the way of CPU when FPGA came on-line ... scrapped my GPU expansion plans out of sheer nerves last week ...

EDIT: here it is
http://forum.bitcoin.org/index.php?topic=6903.msg101253#msg101253

   
Re: Bitcoin mining more profitable than January.
01 May 2011, 13:15:05
   Reply with quote Edit message Delete message  #7
Bitcoin minting ... i prefer that term to the mining term i think ... with a pool it is more like minting, while going solo it is more like mining (due to the variance differences).

What the current crop of GPU miners need to be aware of is the fpga threat on the horizon. As the CPU miners were out-moded when capital values involved in bitcoin became high enough, then so to maybe GPUs become out-moded when capital values of bitcoin minting gets to high enough levels to justify fpga investment. It has been more about capital equipment cost (silicon chip) than electricity.

Some banks and trading houses already have fpga hardware, some may even have it sitting around idle after the financial crash, so some programming and electricity and they are in business bitcoin minting without significant capital expenditure ... (ps PM me if you would like to know how to do this, if you can pay in BTC).

Edit: big GPU clusters may be around for a long time too, hard to know

[kr105]

So I'm gonna have a guess that it was an existing financial entity of some sort ... which means they will have an incentive to set-up a decent bitcoin exchange with easy interface with existing currency markets to go along with their mining operation. Basically a 'rogue' form the current oligopolies is breaking ranks and going to "have a go". If this is the case, it is great news.

Okay, yes. But it's very, very hard good news to handle for those of us who just dropped a couple grand on mining equipment only a few days ago. I want to at least make back my investment before FPGA mining makes it all worthless! 

I second that. But AFAIK this will not be massive anytime soon, so maybe we still have time to recover our investement

[SgtSpike]

I actually view it as a GOOD thing if an FPGA cluster was what the big hashing increase was due to.  Hopefully, that means that the difficulty rate increase will slow down, unless/until more FPGA clusters make their way online.

If it's just due to a bunch of us buying more GPU's, then there's no reason that the difficulty increases won't continue for quite a while.

[MoonShadow]

Apologies but no more development information will be posted.  I've been offered a 25% share from someone that owns 2 FPGA clusters.  If you haven't seen that type of hardware before think a 156 FPGAs per machine.

Okay, so the most likely existing FPGA user with this kind of hardware is either a crypto set-up or banking/financial house that used it for algo trading.

Perhaps a financial institution breaking ranks, but it could also be the CIA looking to test their intentions for bitcoin on idle hardware.

[bulanula]

Looks like the GPU is going the way of Barney the dinosaur LOL !

Bring on the FPGA fools ! I never bothered with GPUs anyway !

[Ian Maxwell]

So the age of the GPU is coming to a close? We knew it was coming, I just thought it would take longer.

Ah well... I went in hoping for a free video card, and I ended up making a pretty solid profit on top of it. I can't complain. Maybe I'll start learning about ASICs or something... or maybe I'll just enjoy the chance to play Portal 2 without feeling like I'm burning money.

[gusti]

I would like someone to show the calculations proving that fpa or asic mining is actually more cost effective than a decent GPU.

[bulanula]

I would like someone to show the calculations proving that fpa or asic mining is actually more cost effective than a decent GPU.

Indeed me too but cannot ignore how CPU is dead even if I have free electricity etc.

Price is no object here it seems so the more power a technology has the better. Eg cost of this don't matter. If it can take out gpu in speed then gpu dead etc.

[gusti]

I would like someone to show the calculations proving that fpa or asic mining is actually more cost effective than a decent GPU.

Indeed me too but cannot ignore how CPU is dead even if I have free electricity etc.

The reason is that CPU is not cost feasible anymore, even with free electricity. 

[marcus_of_augustus]

What we need now are some bounties posted to get FPGA technology open sourced ... links to good cards and s/ware and set-up guides ... exactly as was done for GPU back in Jan/Feb.

If the bounties are posted in BTC then the cost to the outside world becomes slightly less relevant since the the current hardware has probably paid for itself, BTCwise. It is just another demonstration that BTC is actually backed by the hardware/software and electricity going into keeping the network competitive, secure.

FPGA miner bounty, who's up for it?

[gigabytecoin]

What we need now are some bounties posted to get FPGA technology open sourced ... links to good cards and s/ware and set-up guides ... exactly as was done for GPU back in Jan/Feb.

If the bounties are posted in BTC then the cost to the outside world becomes slightly less relevant since the the current hardware has probably paid for itself, BTCwise. It is just another demonstration that BTC is actually backed by the hardware/software and electricity going into keeping the network competitive, secure.

FPGA miner bounty, who's up for it?

I would rather a bounty be created than somebody do it on their own time/for themselves. Message me if there is one ever going..

[Cheeseman]

Why is everyone getting upset over FPGAs? They're not a threat at all, and when they do become a "threat", they will be freely marketed to miners.

The best price (academic) I can get on an Altera 115K is $330. It is theorized they can do about 100mhash/s (currently only 80mhash). That comes out to $3.30 per mhash, compared to the $0.50 per mhash of a 5870. Thus it would take over six times as long to breakeven on the hardware, which increases variability (risk) by a ton. Not only that, but FPGAs have little resale value. I can easily turn around and sell my 5870/5970 for what I purchased it for.

A 1 thash/s FPGA setup would cost roughly $2,000,000 to $3,000,000 depending on if a custom board was made. Doesn't make much since, especially considering that bringing that much power to the network would raise difficulty, and decrease your profit. Might as well just pump and dump the currency itself to see a better return faster.

The ONLY advantage of an FPGA currently is its long term sustainability. Difficulty will get to a point where GPU miners can't mine for a profit after electricity, so FPGAs will have to come in. Or the huge decrease in hashing power after that point will lower difficulty and make it profitable for the GPU guys again.

[exahash]

Apologies but no more development information will be posted.  I've been offered a 25% share from someone that owns 2 FPGA clusters.  If you haven't seen that type of hardware before think a 156 FPGAs per machine.

Okay, so the most likely existing FPGA user with this kind of hardware is either a crypto set-up or banking/financial house that used it for algo trading.

Funny that the fpga conspiracy theory jumps right to the financial sector.  I know one shop that does a lot with CUDA and heard that CUDA's the reason Amazon went with Nvidia cards.

Had a consult with an engineer today and the short of it is that going out and buying the fpga's just to do mining is unlikely to be profitable.  ASICs will cost even more, but if someone really wanted to throw 7-8 figures (USD) at it, they could exceed the current hash rate of the entire network.

The best way he came up with is to do what cypherfox is doing - find someone who already has a ton of idle capacity (making acquisition of the fpgas a sunk cost), and partner up.  His connections were not private sector, so I have no leads on idle capacity at the moment.

[Dhomochevsky]

I'll have to agree with Cheeseman. I figure GPU mining will still be profitable at least 2 months from now (provided the giant space monolith on the Moon doesn't figure mining for BTC is a fun pastime) at least profitable enough for people to break even... What we will see though is a slight settling of the difficulty rate up to a level where is worth it but just barely, increasing the time for a ROI. Then the difficulty will drop slightly for a couple of cycles, then rise again to those levels. When a situation like this is reached, that's when FPGA/ASIC mining will start being used more and more and we'll have miners slowly adopting custom boards. Then we'll have a mix of GPU/FPGA mining. This will make the difficulty levels creep up extremely slow.

The reason I think at some point the difficulty will stagnate is that, while new people will find out about bitcoin and try to join in on the mining craze, they'll see it's not worth doing it without some major hassle. So with a slight variation, the difficulty rate will be settled at some point and will only rise when new technology is introduced in the network. We will have reached the limits of our available computing technology and we'll have to wait for it to catch up.

The stagnation of the mining rate, coupled with increased popularity will mean the bitcoin's value will go up. This will invariably change the profitability rates and will get the ball rolling a little more, albeit slowly.

What I'm saying is that we'll reach a point where the profitability will be much smaller than today, and it will be linear relative to the difficulty climbs. But GPU mining will not cease to be profitable too soon, although yes, profits will probably be ridiculously small eventually. Also, if the community FPGA thing flies, custom boards will slowly start replacing GPU setups, but make no mistake, it won't be as fast as the GPU replacing the CPU.

Before anyone jumps on my head, yes, investing in BTC right now is a much more sound strategy than buying hardware and setting it up for mining. Following the general trend we've seen since a year ago, the bitcoin will only grow in popularity in time. This means more transactions, this means more circulation, more exposure and by extension more value per BTC. But mining will always have a niche and people won't stop doing it, it's a nice natural mechanism forming around the network.

Of course, everything I said here could be moot provided some higher power decides to fuck around with the network, but I don't see the governments trying to curb the Bitcoin craze while they don't see it as a danger to the status quo. At least not until next year.

[edit] Perhaps I need to give some clarification on the "next year" thing. How should I put it... You see, when animals are in the mating season they do all kinds of crazy things. Well, next year is an election year for a couple of countries I know; it's mating season for the politicians and you see, politicians tend to do all kinds of crazy things when in mating season. I'm not saying an apocalyptic crackdown is inevitable, I'm just saying the more popular Bitcoin becomes, the more it's bound to raise some eyebrows. This can also prove to be a good thing (governments attacking Bitcoin). I'd laugh my ass of once the Streisand effect goes in full gear over something like this.

[cypherf0x]

FPGAs are only cost effective if you already have them for other purposes.