X6500 Custom FPGA Miner

[BombaUcigasa]

sorry, am i missing something? who would pay $600+ for 180MH???

Well you can hook it up to a 30W solar panel which costs another $200...  and you get free bitcoins forever? Well 0.1 BTC per day...

Cool, so you will pay it off in 2.5 years.

Or keep going, or sell back the hardware at 25% discount, so it pays off actually in 8 months

[Karmicads]

Not to rain on your parade, but be realistic:

1-Don't count on a 30w consumer solar panel to still power your devices in 3 years, and your batteries will need to be replaced several times over that stretch. 15w for 14 hours is an expensive battery and will deteriorate quickly from frequent deep cycling.

30w? I was thinking more like 130-140w for 4 X6500's (eight FPGU's). That's 10 or 20w more than the total load. So there's a little room for deterioration. I don't expect more than 1.5% deterioration per year. So 15% over a whole decade. Perhaps I will go as high as 150w. The batteries I have to think about some more. Even if it needs a another $500 or $1000 spent on Absorbed Glass Matt batteries, the whole thing is still going to pay for itself several times over, well within the life of the panel. It would pay for itself twice even without batteries and only running in daylight hours. The extra cost of battery backup so it can run 24/7 would be marginal to the initial setup costs. Even allowing for two or three battery replacements over the decade.

2-Unless it's extremely well-placed and on a heliostat, your panel will operate at 100% for a grand total of 0% of the time and will only approach 100% for a small part of the day.

So we need to aim at something more like 85% average over 10 hours. I think that's doable. Also a heliostat might very well be a wise investment. That or solar tracking. As long as the solar panel is reasonably oversize the need for 100% efficiency is averted.

3-Regardless of where you live, it is not sunny 100% of the time. Find out your longitude and latitude and google a solar calculator, enter your info and see what kind of realistic production you can expect.

I'm well aware of this. I've been looking into solar systems for quite a while now. I live in South Australia we have an excellent solar yield.  

4-You still gotta run a computer.

That's a negligible overhead in terms of running costs. I'm still not convinced that an on site processor is necessary for a network connected unit.  Even still, I cant imagine needing anything as powerful as the processor in a typical palmtop or even an embedded system. If the software can be run over the network, all the better. The point is you don't need a huge power hungry monster that runs on mains power.

Small scale solar is cost prohibitive.

And that's why everybody's doing it if they can afford the upfront cost? C'mon now. Haven't you heard? People are putting $5-15,000 dollar units on the family home, that pay for themselves in under 5 years, cut their electricity bill down to a small fraction and sometimes even generate a small profit. Even selling excess power back to the grid is profitable. If buying power from the grid to mine bitcoins (with power hungry GPU's) is profitable, then cutting out the middle man by turning solar into bitcoins without inverters/grid-power at a tenth of the wattage, must easily be profitable.

[Karmicads]

How about this, we'll be much more willing to accept new orders if you're planning to pay with BTC!

What I think I would have preferred, is a price fixed in BTC at the time of order (at least the maximum), rather than the price being set in USD terms. There's no way to tell how much the order will cost in BTC, between making the commitment and the time to pay the balance. It's a little difficult for anybody to deal with the fluctuating value of BTC, but if the agreed exchange rate is set at the time of commitment, then we'd have a good idea of the most BTC we would have to find to cover the balance. Just a little something to consider, if you preffer payment in bitcoin. 

[fizzisist]

How about this, we'll be much more willing to accept new orders if you're planning to pay with BTC!

What I think I would have preferred, is a price fixed in BTC at the time of order (at least the maximum), rather than the price being set in USD terms. There's no way to tell how much the order will cost in BTC, between making the commitment and the time to pay the balance. It's a little difficult for anybody to deal with the fluctuating value of BTC, but if the agreed exchange rate is set at the time of commitment, then we'd have a good idea of the most BTC we would have to find to cover the balance. Just a little something to consider, if you preffer payment in bitcoin. 

Cablesaurus uses a script that automatically adjusts the BTC prices in terms of the latest Mt. Gox price (or weighted average, I'm not sure). This will give him a predictable amount of money for each product he sells.

I know that can be a little frustrating, but remember, we're paying for these boards in USD. What happens if we set the price in BTC today and the exchange rate goes down next week? We could easily end up giving the boards away at a loss! It's a shame, but Bitcoin really isn't the currency here, only the method of transfer.

If you're really worried about the fluctuations, you can exchange that amount of BTC for USD right now. When it comes time to pay, convert it back to BTC and you'll have just the right amount. Well, except for the exchange fees.

Sorry!

[Karmicads]

If you're really worried about the fluctuations, you can exchange that amount of BTC for USD right now. When it comes time to pay, convert it back to BTC and you'll have just the right amount. Well, except for the exchange fees.

Now there's a good idea. Why didn't I think of it? Don't answer that.  

Hang on!   I just realized something (halfway through selling my entire BTC stash). What if the price goes up again and I'm holding USD. I'm not going to have "just the right amount", because I'll be burned for selling low (it's pretty low as it is) and buying high.  

I'm feeling a bit tense right now.    Seems I always manage to screw up when I try to do anything with bitcoin on the exchanges.  

EDIT: Er... Strike that. Of course - I just realized the USD price is fixed so however many BTC I get when I buy back in will cover the price. 

[molecular]

FPGA is now searching for lottery ticket...
I've got a Golden Ticket!!
7BD107D0
Hex nonce: 7BD107D0
New nonce hex: 7bd107d0
Original data: 00000001e8162cee2367fac1993d45171558e47140cd2bb5efad0d1b0000077400000000df73099 e56202184f57a5fec25eeb7840d311513aa89894ea24e3e927a172c244e5e48501a096fe3000000 0000000080000000000000000000000000000000000000000000000000000000000000000000000 0000000000080020000
Golden data: 00000001e8162cee2367fac1993d45171558e47140cd2bb5efad0d1b0000077400000000df73099 e56202184f57a5fec25eeb7840d311513aa89894ea24e3e927a172c244e5e48501a096fe3d007d1 7b00000080000000000000000000000000000000000000000000000000000000000000000000000 0000000000080020000
Submitting work ...
Result: {"id":1,"error":null,"result":true}

YESSSS!

I got it working, on LINUX! fucking hell yea!

Li, newMeat, should I write up a little linux guide. It's not been trivial ^^

EDIT: Arghhh! sorry, this is the wrong thread. will crosspost to x5000 thread

[newMeat1]

Haha, you're the man molecular!    

I don't think fpgaminer has even gotten it running on Linux yet (cause he's been really busy lately). Yeah, please write something up... then add it to the wiki, or we can add it for you.
http://fpgamining.com/index.php?title=Main_Page

EDIT- give us a picture here, too, please! It's amazing how small that thing looks.

Here are some FPGA mining articles we've found. Interesting stuff!
http://hackaday.com/2011/08/22/fpga-bitcoin-miner-is-probably-the-most-power-efficient/
http://www.bitcoinminer.com/post/9612343341/all-things-it-fpga-bitcoin-mining

[newMeat1]

Here's a little cost analysis of our board. You guys will find this interesting, I think. Basically it shows that the preorders were enough to "get us over the hump", where the price is really steep. From here on out, there's little gain to be had, until we get to 1,000 units.


It's a good thing we got at least 50 orders, otherwise I don't think we could have delivered for under $620. Like fizzisist was saying, all those little taxes, shipping, and Paypal fees really add up.

[Karmicads]

(Bump) How's the new design going guys?

[Christian Pezza]

Any comment about this? http://www.enusbaum.com/blog/2011/08/23/why-the-idea-of-fpga-bitcoin-mining-is-stupid/

hope it's legit to post the link here

[defxor]

Any comment about this?

Yes, the author's logic circuitry is broken

There's a valid answer posted to the article already that brings up electricity costs. It's not just about the ROI with the difficulty levels of today but at what price level a GPU miner vs an FPGA (or ASIC) miner can keep running their rigs.

Quoting from the author in the comments:

So at the end of the FPGA break even period, you’ll have ONE paid off FPGA cranking out 300 Mhash/sec, or two 6990′s putting out 1.5 Ghash/sec

"Paid off" doesn't negate the cost of running the rig. 750W for the GPUs, 16W for the FPGA (numbers from the article) = 47 times more expensive. 1.5Ghash/s vs 300Mhash/s = 5 times more hashing power.

5/47 = 0.11

So, the GPU solution the author proposes, with reinvesting the profits into more GPUs, is 11% as effective as going down the FPGA route.

(After having written all this I'm somewhat surprised by my own numbers, but I'll post it anyway. I suspect I'll be corrected)

[Karmicads]

Any comment about this? http://www.enusbaum.com/blog/2011/08/23/why-the-idea-of-fpga-bitcoin-mining-is-stupid/

hope it's legit to post the link here

Building himself up to his own level of self importance:

here’s the part where I let everyone in on a little secret: Ready?

[Insert: Dramatic pregnant pause]
 

Are you sure?

 
[Insert: Another dramatic pregnant pause]
 

Ok, but don’t say I didnt warn you:

[OK! I'll try to hold back my excitement]
 

Modern GPU’s are (essentially) FPGA’s
Did I blow your mind? Do you need a minute?

What people seem to forget is that modern GPU’s are essentially FPGA’s in that you’re writing a program that will take advantage of a dedicated computing resource for which the management of allocation and assignment is abstracted out through the hardware.

ATI/AMD cards for example make extensive use of Stream Processors (or as they were once called, “Programmable Shaders“) which allows the equivalent of the FPGA ‘cores’ to be programmed and executed on the GPU.

So what we’re left with is a video card that is also essentially a FPGA that can execute custom programs in a massively parallel environment. And just like Alton Brown, I love a multitasker! So there’s absolutely no need for FPGA miners as they’ll never be cost effective and the alternatives yield much higher returns.

WANKER!! The simple fact is that GPU's DON'T compete NOW with FPGA technology and they are not optimized for  SHA256 hashing but for video preprocessing.  That's WHY they are so inefficient with mining bitcoin compared to the FPGA. Can the author of this article write some code to make a 6990 do (in terms of energy costs) what the Spartan 6 will do? NO?I thought not. If GPU's could do what dedicated bitcoin mining FPGA's could do, then these guys (fpgaminer, fizzisist, newMeat1 et. al.), would have been deploying firmware upgrades instead of the costly (but worthwhile) FPGA based hardware solution. You have to account for difficulty increases and power costs. This guy can't see past his nose.

[spiccioli]

Hi guys,

did someone look at this site: http://www.picocomputing.com ?

Could they have something that helps producing faster/cheaper boards?

Cheers.

spiccioli.

[Vladimir]

Hi guys,

did someone look at this site: http://www.picocomputing.com ?

Could they have something that helps producing faster/cheaper boards?

Cheers.

spiccioli.

I had a chat with them in 2010 before I went GPU route. Their prices were not competitive comparing to GPU's. I doubt they are any better now and surely not any better than home-grown variety of FPGA boards developed by Bitcoin community recently.

Though of course maybe they quoted me a silly price first. Not the price they really ready to take. All this marketing BS such people play all the time.

[fizzisist]

Hi guys,

did someone look at this site: http://www.picocomputing.com ?

Could they have something that helps producing faster/cheaper boards?

I had a chat with them in 2010 before I went GPU route. Their prices were not competitive comparing to GPU's. I doubt they are any better now and surely not any better than home-grown variety of FPGA boards developed by Bitcoin community recently.

Though of course maybe they quoted me a silly price first. Not the price they really ready to take. All this marketing BS such people play all the time.

Yep, I got a quote for the M-502 (2 Spartan 6 LX150s) and it was over 4 times the price of our board... I'm not even sure if that includes the PCIe card that those modules ride on. Certainly not cost competitive, even with most FPGA dev boards when it comes to mining.

WANKER!! The simple fact is that GPU's DON'T compete NOW with FPGA technology and they are not optimized for  SHA256 hashing but for video preprocessing.  That's WHY they are so inefficient with mining bitcoin compared to the FPGA. Can the author of this article write some code to make a 6990 do (in terms of energy costs) what the Spartan 6 will do? NO?I thought not. If GPU's could do what dedicated bitcoin mining FPGA's could do, then these guys (fpgaminer, fizzisist, newMeat1 et. al.), would have been deploying firmware upgrades instead of the costly (but worthwhile) FPGA based hardware solution. You have to account for difficulty increases and power costs. This guy can't see past his nose.

I love your reaction to this, Karmicads. I'm not sure what this guy's trying to get at here. Maybe he's been working on some new firmware for the GPUs that will be much more efficient at mining? I certainly have no idea how to do this.

Also, by the same logic, GPUs are essentially CPUs. Why would you mine with GPUs, then?

We can all see the FPGA more efficient but to me its also more risky.  The GPU's are an asset thats easy to sell for a good price.  good luck selling $620 FPGA boards if bitcoin tanks.  I would MUCH rather have a $620 PC then a $620 FPGA.  The pc is useful even if i chose not to bitcoin.  Or i can sell it for around $620 if bitcoin  is gone.

This is a valid point, but you also have to factor in the depreciation of the GPUs. Furthermore, as we see with the Pico computing example, people are interested in FPGA coprocessors for other cryptography and computational work, and they're willing to pay a lot of money for them. There's a good chance you could resell our boards to people working in that field if Bitcoin did completely collapse.

[Vladimir]

picocomputing.com and their ilk are feasting on governmental, spooky, military and bank funded customers, who do not count every penny, since it is all paid by tax payers one way or another.

I doubt that those providers ever will get prices to a reasonable level for competitive bitcoin mining market.  If they would want to get into this market, their margins will have to shrink significantly.

[newMeat1]

How's the new design going guys?

Pretty well, Karmicads. Today we talked about how we're going to protect the circuit from overcurrent. We might also be adding overvoltage protection. Small details basically. I'm pushing hard to send the board design out on Sunday afternoon, to get our prototype made.

[RandyFolds]

Any comment about this? http://www.enusbaum.com/blog/2011/08/23/why-the-idea-of-fpga-bitcoin-mining-is-stupid/

hope it's legit to post the link here

Building himself up to his own level of self importance:

here’s the part where I let everyone in on a little secret: Ready?

[Insert: Dramatic pregnant pause]
 

Are you sure?

 
[Insert: Another dramatic pregnant pause]
 

Ok, but don’t say I didnt warn you:

[OK! I'll try to hold back my excitement]
 

Modern GPU’s are (essentially) FPGA’s
Did I blow your mind? Do you need a minute?

What people seem to forget is that modern GPU’s are essentially FPGA’s in that you’re writing a program that will take advantage of a dedicated computing resource for which the management of allocation and assignment is abstracted out through the hardware.

ATI/AMD cards for example make extensive use of Stream Processors (or as they were once called, “Programmable Shaders“) which allows the equivalent of the FPGA ‘cores’ to be programmed and executed on the GPU.

So what we’re left with is a video card that is also essentially a FPGA that can execute custom programs in a massively parallel environment. And just like Alton Brown, I love a multitasker! So there’s absolutely no need for FPGA miners as they’ll never be cost effective and the alternatives yield much higher returns.

WANKER!! The simple fact is that GPU's DON'T compete NOW with FPGA technology and they are not optimized for  SHA256 hashing but for video preprocessing.  That's WHY they are so inefficient with mining bitcoin compared to the FPGA. Can the author of this article write some code to make a 6990 do (in terms of energy costs) what the Spartan 6 will do? NO?I thought not. If GPU's could do what dedicated bitcoin mining FPGA's could do, then these guys (fpgaminer, fizzisist, newMeat1 et. al.), would have been deploying firmware upgrades instead of the costly (but worthwhile) FPGA based hardware solution. You have to account for difficulty increases and power costs. This guy can't see past his nose.

I don't know why everyone gets so uppity about stuff like this. It is an interesting application of fringe technology. It may or may not pay off...

I see too much crap about Moore's law and the difficulty/price point at which GPU mining becomes inefficient. Difficulty has dropped for a month straight. The market is in chaos. The social scene is in complete and utter chaos. Who pretends they can read the future in such tumultuous times?

I think the author makes an excellent point, in his 2-year comparison, or maybe that was a commenter. They all seem to be named Eric.

[Karmicads]

How's the new design going guys?

Pretty well, Karmicads. Today we talked about how we're going to protect the circuit from overcurrent. We might also be adding overvoltage protection. Small details basically. I'm pushing hard to send the board design out on Sunday afternoon, to get our prototype made.

Noice. Hows about a reverse polarity diode. Only be a few cents extra. I assume it'd fry something if the supply polarity were hooked up backwards.

[ElectricMucus]

What seems to be a real bummer here and at the competition is that these fpgas cannot be developed for with the free xilinx webpack.
That's real bad for anyone interested in hdl design as well.