Bitcoin Forum

I would like to do the long-term mining via these, to save on power.

Anyone knows if some are coming to market?

Have you looked at these: http://www.picocomputing.com/ex_series.html

I'm working on a 1st prototype right now. I'm glad there's interest  :). There are more advantages than just power savings: FPGAs will also be quiet and you don't need a motherboard slot. The only hardware you need is a USB port. And the speed will go up in the future (80 Mhash/s was just a first attempt, by fpgaminer). He's working on putting a second core on a Cyclone IV right now. They get 80 Mhash/s with about 10 watts-- a pretty damn good power ratio

Anyways, next weekend, I'll have all the parts to try the board out. If it works, I'll start selling them on Sunday for around $450. That's a $150 cut off the Terasic development board price. If it doesn't work, I'll hang my head in shame and go back to the drawing board. Haha

Let me also mention- there's no fan. It's a passive heat sink

Quite expensive per MH/s. The market is so fickle and volatile right now, I don't know that there'd be any interest in FPGA beyond the cool / hobby factor unless costs were comparable on a 3 month timeline, including the power 5-10x power savings. As I calculate it, at current BTCUSD exchange, it would take 300 days to break even for $450 running 80 MH/s @ 10W.

Is there any probability that a 400-800 MH/s unit could be produced on a small scale for $300-400? I don't see it happening. Cool idea though.

But you need to consider the other hardware costs. I mean, you could buy an old laptop on eBay for $20. As long as it has a USB port (or several...) and an internet connection, you're good to go. I've already found cheap suppliers for the power supply and cables, so those would be included in the package.

And if fpgaminer can get the speed up to, say, 120 Mhash/s, it starts to get really competitive...

Can I see your maths?

Here's what I see..

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

Um I haven't done any math yet. I'm just thinking in concepts. Long term, obviously, efficiency is gonna trump raw hashing power. The less efficient miners are gonna start dropping out.

Can I copy what you came up with? That's good.

I would be interested in multi-fpga boards if you have plans for that!

I think that difficulty is going to exceed the capability of any FPGA units going forward, and there isn't a feasible way to beat the ongoing and increasing difficulty using FPGA unless you can do 5 cores x ~80 MH/s for around $200 - $300. If there are existing multi core, commodity FPGA boards that could be relatively easily re-purposed for Bitcoin mining, well.. Then you have something. If those boards exist, the next step is to find out if you can get them at a discount on volume. You have to make money, right?

Electricity costs can definitely impact profitability, but to be competitive with commodity GPU mining, the scenario mentioned above would need to happen before there is any possibility of it being an alternative or successor to existing methods.

to save power? are you serious??

do you not understand that the computation power in bitcoin is used as the matter to compensate for growing computational capacity?

definitely I was not clear enough.

ok to say it in other words. It does not matter how much computation power you got per electricity consumption. eventually, to catch up, every other miner will get the same rig as you and then the only way to compete will be to increase the number of those rigs you got. hence the power consumption will grow to previous level because of rapid difficulty growth that is used to consume any computational power you can throw to it. No matter how much you optimize, bitcoin generation will always consume as much power as miners can put their hands on.

this is somewhat like the arms race... and it's just as pointless. but it can't be stopped. this is an intrinsic flaw of the bitcoin design unfortunately.

Yes, eventually low power, high density miners will win and GPU miners will lose out.  How many 4x6990 power hungry mining rigs can one put in a normal house?  how to deal with the heat?  the noise? 

How many cool, efficient and silent FPGA boards can be run? 

There will be a new technology and a tipping point to make GPU obsolete.  It may not be today, it may not be fpga, maybe it is ASIC?  but the day will come and power and efficiency will be a major factor.

Agreed. However, whatever that solution is, it's going to have to be a widely available technology, very likely with an alternate use (like ATI GPUs) or else there will be no other option. Also, unless the BTCUSD exchange rate stabilizes, and more merchants start accepting BTC, there's not much likelihood that any FPGA or ASIC commercialization projects will ever exist for the Bitcoin community. It will be limited to hobbyists.

You are wrong. When the electrical cost exceeds the bitcoin value, many will stop mining, causing a difficulty drop. The price/difficulty will eventually settle at a point where the average miner breaks even.

FPGA and ASIC miners will be the only ones making a profit.

ASIC is the way to go.

FPGA will always have a steep price point, cheap ASIC will be the one to make a dent in GPU's hashes.

What do you think the cost to develop a commercial FPGA and/or ASIC solution would cost? Perhaps some engineers should put their heads together and come up with realistic specs, an open source design, and the software to run it. I don't see that happening, and unless someone decides to invest a few hundred thousand dollars into specialized equipment, and the time to sell specialty FPGA and/or ASIC Mining devices, I don't think it's going to happen.

Where's the incentive for anyone to go down this road right now? In 3 months? Even if difficulty drops (it's not going to), there's too many people who will be willing to deal with the heat of GPU mining, and not enough people unified enough to develop an alternative that's viable for the Bitcoin community, at least as far as MH/s per $ goes.

Good discussion on the economics of FPGA/ASIC mining...

http://forum.bitcoin.org/?topic=2362.0 (http://forum.bitcoin.org/?topic=2362.0)

Not true.  A dedicated mining board with ASICs will sell like hot cakes.  People will make huge arrays and farms and provide the mining power that the bitcoin system needs.  

It will be a business decision to buy them, they will pay for themselves in X weeks/months then they will have a lifetime of x months or years to make money.  After that they can be thrown away.  

Many are treating GPUs the same way now.  If you bought 50 6990 cards today, and they paid for themselves in 2 months.  Would you care about selling them on ebay in 2 years for $50?  Or would you just run them till they die or become obsolete (or no longer profitable)?

which is exactly my point.

what prevents those FPGA/ASIC miners to get more of those things to increase their profits? And so the race begins.

Of course if only the chosen few will own those things they can have pretty small total power consumption but does that not defeat the purpose of a bitcoin as a system that is supposed to be p2p, decentralized system where everybody is supposed to take part in transaction processing? those chosen few with FPGA/ASICs will be the real centers of all transaction processing.

Or if general public gets to buy those FPGA/ASIC then again the one who owns the most of them, wins.

Unless of course the price of generated bitcoins no longer pays for the electricity spent... in which case we get decline in mining, difficulty adjustment, resume in mining, cycle repeats.

Once a fully working and bug free FPGA design is ready it will take this to move to full ASIC production.  (from what I understand about it so far).

cost - $200k - $1.5MM
time - 6 - 18 months

So you have to find an investor who believes in Bitcoin enough to trust your estimates for difficulty in 6-18 months, and you'd have to ensure that you weren't late to market with a unit or units that were already behind the curve, right?

If Bitcoin continues down the road it's on, then yes - there will be dedicated ASIC bitmining units sold commercially. However, it's a long road from here to there.

Yes, it would be a risky investment of capital and time for sure.   

I'm not an FPGA person, but have a few questions that might help get a trajectory on all this...
Are logic cells and LUT's equivalent in quanity? That is, do they come one for one within an FPGA board?
What number of logic cells do you think is need to acquire 100MH/s?

Finally...

Do you have any idea's on the actual costs of these below items? To include the super cluster?
I can not find any pricing information:

http://www.picocomputing.com/ex_series.html

http://www.linera.com.tr/temsilcilikler/pico-computing

BTW: I think you made a very valid point earlier about the fact that even though the boards would cost more, that you could run them with less hardware. Many USB ports can be run from one computer (even with a GPU or two on the MoBo  :D). The extra cost may be saved, in part at least, by the non need to purchase extra fans, larger power supplies, mother boards....etc... not to mention the power savings with FPGA. A comprehensive cost analysis would be beneficial. Knowing what hidden costs might exist in scaling would be interesting. For example, any costs to modify electrical or coolings systems for expanding to more GPU rigs.

Yeah, ASIC would probably be pretty steep. I don't have any experience with it, but I imagine...

FPGA is really cheap to get going, on the other hand, if you have the skills. The first prototype is gonna cost me $500. I'm soldering it in a $30 dollar toaster oven  :)  And I plan on selling small batches, maybe 5-6 a week, just to make a modest profit. Someday maybe it will be bigger than that, we'll see.

I may be looking for some help with this, down the road. This might be too early to even think about, but I'm gonna be busy soon.  So, if it works and takes off, I could use some assembly help

What are the specs of your first prototype? What do you see the specs of your 5-6 a week being, and at what cost? Would a larger upfront purchase of parts yield you a significant cost savings?

Regarding being behind the curve. If the design was compact, efficient, and modular for scalability, then it seems it wouldn't be a major concern about subsequent designs or technology (except maybe quantum computers  :P).  As I understand it, the FPGA circuits would all be executing the same process in many parallel paths within the circuitry anyway. Scaling it simply by some modular means seems most attainable from a layman's viewpoint.

You can also get FPGAs mounted on boards with all kinds of interface options, USB and whatnot, ready-made from the manufacturers and those are really easy to program if you want to do it yourself. They're more expensive, though, since you're paying for all the extra hardware as well as the FPGA, but updating and sharing the algorithm would be a walk in the park. Should be great for people who already own an FPGA board, or people who want to get one anyway for the reprogrammability.

Hi,

has already someone spent some time to create a list with the various FPGA boards that are available, the cost of the board and the maximum hash hash rate?

Has anyone started a company to professionally produce these? Not like a huge mega venture but just something small and professional?

See here: https://bitcointalk.org/index.php?topic=37904.0

I don't think that's going to happen very soon... Individuals may order batches of even tens of mining boards but a whole company? That sounds too risky to me, atleast.

The market is to small for a dedicatet commercial "FPGA mining board".