Topic: 1GH/s, 20w, $700 (was $500) — Butterflylabs, is it for real? (Part 2) - page 31. (Read 146909 times)

Who's got their tracking info? Two days until they totally fucking blow another deadline...amateurs.

I was wondering when you'd come out!

LOL. At least we got the chips data !

So, it's safe to assume 4-6 weeks?

What a bullshit update. Chinese New Year's, my ass...

What's funny is that the information in this forum has consistently been more up-to-date than the 'updates' on BFL's site. And that's thanks to all those who have posted BFL's answers to their questions, as well as photos provided to customers directly from BFL.

I think those with orders are following the BFL site updates closely, but it has become clear that this forum has become a more relevant source of up-to-date information. Regardless of the occasional troll, fud, or off-topic post.

Keep up the information flow everyone; this coming week should see the first shipment assuming nothing 'bad' has turned up over at BFL during assembly and test.

January production update from BFL

http://www.butterflylabs.com/january-production-update/

They might just rougly estimate the power draw.
It seems as if they took information from this document:
http://www.ti.com/general/docs/gencontent.tsp?contentId=70774
-x

Yes, in fact that's an easy mistake to make.

Case in point: The CEO (and probably sole proprietor) of ZTEX did.

Before the 1.15x module, with its 8 Amp core voltage supply, came the 1.15d module, and the originally recommended supply for it
http://www.ztex.de/usb-fpga-1/pwr-1.0.e.html only sported a 3 Amp core voltage supply!

He found out the hard way that all these unrolled loops of SHA-2 cause something like 50% of all flip-flops on the FPGA to switch simultaneously, and
thus blowing even the most conservative power estimations out of the water.

Now he has discontinued the power supply module 1.0 and is instead offering the power supply module 1.1
http://www.ztex.de/usb-fpga-1/pwr-1.1.e.html with an 8 Amp core voltage supply, i.e. the same number of Amps that module 1.15x has.

Q.E.D.

God, this is the stupidest thing I ever heard.  WTF? Everyone knows Unicorns only come in white.  DUh.   

One question.  Do you think it is possible for anyone with the experience necessary to design a custom FGPA board to get the power draw so wrong like BFL claims they did.

I mean 1GH @ 20W on a 65nm FPGA?  My albeit indirect knowledge says that isn't just good it is downright impossible.  I don't care what chip you use you aren't getting that level of performance from a 65nm tech.  They should have known it was impossible right?  I mean it would be like Toyota changing one component in their Prius and the simulated/tested gas mileage goes from 40mpg to 300 mpg.  It is obvious that is impossible.  Toyota wouldn't be announcing the 2013 Prius getting 300 mpg.

It would for me.  Nobody buys FPGA retail.  Nobody except someone who needs a SINGLE chip for a prototype.  Wholesale FPGA go for 30% to 50% off at that is normal business.  Deeper pricecuts are certainly possible when the company is looking to dump them.  

Look the Statix III is a 65nm chip.  Alterra isn't even making it anymore.  They want to sell Stratix IV and soon they will need to sell Stratix V.  They want that inventory gone.  Gone from their website, gone from wholesalers, gone from retail outlets.  when Stratix V is in full production it becomes the high end part and the Statix IV becomes the value segment.  The Stratix III is just a third wheel.

Using a product anyone can relate to a top of the line Pentium 4 was $400 chip at launch.  If today someone found a tray of 1000 of them would it be impossible to believe they would unload them for say $20,000 ($20 ea).  95% off.  Holy shit.  Unbelievable.   No just business as usual.

While getting a last gen chip cheap is no guarantee they will deliver on their promises, it at least makes their claims plausible.

Marketing nonsense like "hybrid blend of FPGA and ASIC technology" just fueled speculation on sASIC but the performance, voltage, timeline, and issues all pointed to it not being an sASIC.  On the other hand their claim of 1GH @ 20W is not possible on a 65nm FPGA at any price.  I don't care if you want to pay $1000 per MH you aren't getting 50MH/W on a 65nm FPGA.

So you take everything the company did say (in double speak and half truths) and it becomes impossible and fraud/scam starts looking more probable.

I'm not sure that being more open here would have helped BFL all that much. To many that have already posted, this product has been (and perhaps still is) nothing more than vapourware, or the 'magic unicorn'. Being more open wouldn't have changed the 'photos or it didn't happen' sentiments floating around here. BTW, I am not in that camp.

One powerful strategic advantage for BFL is that they could announce the imminent availability of a 1Ghps @ 20W FPGA box. Surely that would scare the living daylights out of any existing FPGA competitor. And not only the ones already shipping product; perhaps more importantly for BFL, it would scare anyone even THINKING about designing/developing a new FPGA design. The designers familiar with FPGA designs in general, and bitcoin mining requirements specifically, would be left scratching their collective heads how such performance/price numbers are even remotely possible.

For BFL, witholding these details is maintains their competitive advantage. And was (and continues to be) in their best interests to keep public details to a minimum, so that FUD would hang over any potential competitor contemplating a similar product.

Providing the chip details to the public (i.e. from us, the forum members) doesn't really help BFL. So what if we know what chip they are using? Does it matter? What matters is that they get the claimed performance/power/price. Granted, if they told us they are using a Stratix III EP3SL200 or 260 we could at least be confident that the performance numbers are kosher. However, knowing that these are $4k chips, there would be disbelief that they could somehow sell a $10k board for under $1k.

So how would that help their credibility with us? It wouldn't. It would still be a 'magic unicorn' to the same group of people here, only a different colored one. The only thing that would dispell the FUD for this sometimes-vocal group of forum members would be product shipments and subsequent test results.

I don't think they got factory left overs. I think they scooped these up from an abandoned project. That would explain why the dies have no info.

Solid analysis. 

Which makes all the cloak and dagger and (propreitary blend of FPGA and ASIC technology) so stupid.

The "magic" is the low price last gen FPGA.   Hell BFL could provide a detailed schematic and nobody could copy them.  Well they could except the boards would cost $5K to $10K each.

Had BFL been a little more open with the tech, issues they are having, and solutions to rectify it they could have built some real trust with the community.

Part of my EP3SL150 design:
http://i.imgur.com/FXfnN.png
Part of the BFL sandpapered "bitcoin processor" design:
http://i.imgur.com/ejws0.png
The BFL "bitcoin processor" is Altera Stratix III of unknown size.
The EP3SL150 FPGA configuration bitstream is around 5MB long.
On their board is 8MB flash.
EP3SL150 uncompressed bitstream size is 47 Mbit.
EP3SL200 and 260 is 93 Mbit, which is 11.6 MB.
It doesn't fit into 8MB flash uncompressed, but it may
fit into it with compression.

So my guess is, that they use EP3SL200 or 260.
It is definitely possible to put 3 fully unrolled pipelines into the 260
(at 150MH/s each) and, with some tuning, gain the advertised speed.

To me, the case is closed.
BFL just monetizes their access to chips at prices unavailable to anyone else.

Now, why they just don't resell these $4k+ chips on the market??

-x

Btw, thanks for this image:

That was not the customer base I had in mind. I was thinking more along the lines of small businesses looking to fab 1k+ for resale. That was not at all clear from my post though.

This idea comes up a lot but I don't think  people realize how difficult it is to do manually and how trivially cheap it is to do in bulk via automated assembly house.

For a board that sized if you had 1000 boards built at once you are talking $10 maybe $15 per board and that includes stuff like xray verifcation of solder joints.  Assembly houses often will source all the "mundane" components.  So you just give them the boards, the FPGA, and 2-4 weeks later you have 1000 finished units.

If you gave people the parts the failure rate on a dual FPGA board is going to be 10% or higher.  Of course the company (any company) will get the negative word of mouth from all the people who destroyed their $600 toys.  Trying to do assembly by hand to save money would be like buying a car in parts and assembling it by hand without the proper tools.

I can't speak for anyone else, but it should be apparent from what I have said here and in other posts that I would not be capable of doing so. And most of the guys around here that can are more than capable of designing their own. I think the catch is going to be that even if we could reverse it and it was more prudent to do so than to build our own, that we will not find the same discount on the chips they are using.

And your idea is actually not bad. Design the PCB and Core and then just sell that and/or the chips and skip all the manu costs.

Yeah, I kinda figured that one out but it still amazes me the insisting people on here that want to copy BFL's design

more profit
the reason it was cheap is probably because nobody wants it anymore
except miners