Topic: 1GH/s, 20w, $500 — Butterflylabs, is it a scam? - page 21. (Read 123112 times)

You are right on not being able to extract the bitstream because there is nothing to extract.  Unlike an FPGA which is field programmable (and thus you can steal the "program" an sASIC is mask programmable.  Unless someone breaks into Altera labs or uses some xray analysis to reverse engineer the design you aren't getting the "program".

But dumping open FPA onto a hadcopy would be horrible.  FPGA miner  has the worst performance of all the mining bitstreams out there.  Also it is optimized for 150K LUTs and Altera makes no hardcopy that small.  Just dumping it on a larger chip wouldn't get you very good performance per mm^2 of die space.  It would be very inefficient.  Maybe you brute force some decent performance by simply getting high enough clock but you end up leaving a lot of performance potential behind.

To do it right you need to buy some high end FPGA that matches the specs of the HardCopy you intend to use.  You may even want to buy multiple high end FPGA to try various designs.  Then you need some smart people like rph or ztek or outside telent.  Pay them to take their designs (already higher hashrates than the open FPGA project) and optimize it for the larger LUT count and more efficient routing of the higher end FPGA.  Spend a couple months getting the most efficient design possible.

Only once you have a high performance FPGA design which makes the characteristics of the Hardcopy you intend to use exactly do you burn it.  Contract w/ Altera and have a batch 0f 10K hardcopy sASICS burned with a mask derivted from your optimized FPGA bitstream.  Maybe make the design opensource and raise money via pre-orders to pay for the mask and initial run.

What about it would be low risk?  Even if you consider it low risk convincing some VC types who have never heard of Bitcoin much less the market for mining hardware is quite a sell.  Of course there is always the risk someone else releases an even more optimized design (possibly on a larger run) and the value of your unsold chips plummet.

I doubt you can do that; but even if you could, if your goal is only bitcoin mining, you would probably be better served by taking open fpgaminer and 'hardcopying' that. Whatever BFL did to make their product more flexible is almost certainly at the expense of speed and power efficiency.

Honestly, if I had serious doubts about BFL's claims, Id be talking to some people right now to make a "hardcopy" for bitcoin mining.  (Oh, and then Id set up a rubbish website and accept preorders and bait goat in to a bet to finance half my production costs!)

If the numbers I hear left and right are only ball park accurate, its really a no-brainer and Im surprised some of the FPGA savvy people here are not investigating this course, or at least not talking about it. Unless you are wealthy, it will require VC funding, but that shouldnt be very hard as this is a really low risk venture.

I think they met in a restaurant  . No wonder they were unable to test it.

Thanks a lot for the pictures Inaba! It's very nice to see some progress on this.
I glad to see it working in Linux. I plan on using Linux as well, hopefully I can get the driver to compile on ARM so I can get a ~25 W mining system.

What was the sound level from the fan like? I live in a one room apartment, so I hope the noise level isn't too disturbing.

@P4man: regarding the chips on the board. One of them is an Atmel AT32UC3A1128 micro-controller.

  So, Inaba, when are you going to be getting a unit for testing?  How come they did not leave you with one last night?  *paces*

  hehe, I wonder how long it will take a few of the guys floating around here to emulate the bitstream off of one of these.  Of course even if someone did it will not make getting their own hardcopy any cheaper. But will be a neat project, none the less.

I already did  They declined to answer.  Like I said, it seems a little silly since that knowledge will become public, assuming they hit their target ship dates (and I have no reason to believe they won't at this point), shortly after the first unit ships.  But I think it's a fair restriction given product and their willingness to be open about everything else.

  Different lists. This is the EAR list which most all, ready to use electronics fall under. http://www.bis.doc.gov/licensing/exportingbasics.htm That makes me wonder if thats the number they etched off the board...??

  But, in order to even submit for SHS they would need to have all of the DOJ stuff in order..  In itself it really doesn't mean much more than they are not new to selling electronics that may be available internationaly.

  Edit; CMVP list here(I cannot link directly to the current SHS list);http://csrc.nist.gov/groups/STM/cmvp/documents/140-1/1401val2011.htm
          Note, this is a list of ALL items subimited for validation and are a mix of software, hardware and different techs. Obviously only a very small fraction are suitable or directly designed for Sha-2 exclusively

Not sure what the SHS list is.. but this is on their website:

5. Do your products fall under US encryption export control classification?

Yes, the BitForce SHA256 processor is properly classified under the US Bureau of Industry and Security export control.  Our export control restriction prohibits the export of our products to AT designated countries as specified by the US Department of Commerce.  If you need clarification on your ability to purchase our products, please contact us for review

Is that what you meant?

Inaba can you just ask them whats underneath the heatsink? FPGA, Structured asic or asic?
They may or may not answer it, but it doesnt hurt asking.

  I'd love to see them as well, though they more than likely are blank or just have batch numbers and dates on them if they are custom hardcopies..

  Aye, I agree on the 'not taking the easy approach'  If you look at the current product info on their site for the Bitforce one, there is a 'note' there directed towards customers who will be using this for packet verification, " Note:  Members of our packet verification product program will receive a different firmware package".  That's what I was trying to point towards when I linked the NIST stuff a few pages back. Boeing(which I can't link to) and a slew of other companies have been piling their devices onto the SHS list this year in massive numbers.  It was correctly pointed out that message digesting or 'one way' hashing of docs, etc does not require much processing power. But, doing it at the packet/fragment level for identing the source does. Boeing at the very least, and I would assume more have been working on securing communications at the packet lvl using sha-2 since I was last with them. About 7 years ago...

  Which brings me back to my question for them and to expand on it. "Are you guys gov contractors?" "Is your device or other device you have access to on the SHS list?"


  Still anxious to see if this thing actualy hashes as advertised.

No, I won't have the opportunity to remove the heat sinks to look at or take pictures of the chips underneath.  I guess the shipping units will have to be used for that.  While I agree this is kind of a silly stipulation (Because that will be done with the first one off the line), I believe it stems from caution and not intent to conceal for any deceptive reasons.

That said, yes I will get better pictures of the visible chips once I can get it in better lighting and I get can a proper tripod setup for macro shots.

I know that this is hard to imagine.

Maybe, the BFL people are NOT scammers.

Maybe, consider the possibility, they simply implemented a technological manufacturing/fabrication technique, namely, structured asic, before the competition did the same.

I am waiting to see results of testing.

[edit] technological technique -- odd phrasing

Yeah these will be MAD profitable if they are real NOT.

Little if any resale value outside of Bitcoin.

If you all buy them the difficulty and price will adjust so no mad gains and 1 block a day just like with GPUs ( when only a few had GPUs they were getting mad BTC but when all got GPUs it wasn't that good anymore ) etc.

So you are telling me its hard to believe a company would invest $200k in this while you would be willing to bet $100k that they did not? ROFL

For the money you claim you were willing to bet, BFL could almost produce those chip! Think about it, even if BFL had intended for this to be a scam, they could have accepted your bet and actually produced 1000 s-asics nearly for free!

Thats how clueless you are.

?
Its not rocketscience.  You just download some opensource fgpaminer for altera, you pick up the phone and call Altera. You send them the code and say you want x1000 Hardcopy chips. You pay their price. You receive s-asics. Im barely simplifying.

As far as "designing" chips go, it really doesnt get any easier. By far the hardest part is what several guys on this board have already done by designing fpga miner apps (not too mention PCBs, software etc). You will have to make some minor adjustments, but the bulk of the work implementing that in a structured asic is what altera does for you. Thats the hole point of a structured asic, its literally a "hard copy" of an FPGA app, and those already exist. It only costs money, no magic involved.

BTW, I suspect BFL didnt take this easy approach, I suspect they made their own FPGA design to allow flexibility for other applications besides bitcoin mining, but you could do it like that.

Next time you may want to read up before offering $100.000 bets.

BTW, found some info on timings for Hardcopy:
http://www.altera.com/devices/asic/hardcopy-asics/about/hrd-development-methodology.html

From FPGA to working silicon in 7 weeks. No bad.

Yawn. Pray tell Goat, what kind of "failed" chips could one use to achieve 1GH/s @ 20W? And what makes you think s-asic's could not?

I make re-purposed garbage of it...who knows though...?

Why assume $3? At current prices and difficulty it will take 9.5 months
http://bitcoinx.com/profit/index.php?hashrate=1050&cost=600&electricity=0.15&power=20&months=12.0000&decline=1.00&blockcoins=0

Of course difficulty is likely to go up if everyone starts buying these boxes, and BTC price, well who knows.

IMO if no one else comes out with a substantially better product in the next 12-18 months, you have a fair chance of making a profit, but it its not without risk and does require some assumptions. Unlike GPUs, you will have a hard time reselling these boxes if difficulty explodes due to asics or some other breakthrough.