Topic: 1GH/s, 20w, $500 — Butterflylabs, is it a scam? - page 30. (Read 123112 times)

Im also inclined to take that bet, but it is an interesting idea. Would VPN accelerators be designed for that kind of throughput? Got a link to a  datasheeet for a similar chip like that?

I'll take that bet.   

BFL has already said it's an FPGA, haven't they? Sure, they said something about an ASIC/FPGA combo, but I'm 90% sure the ASIC is the communication IC, not the hasher.

If it were that specific, do you think that would be possible?

OK, I'm officially registering a second bet with myself.

The BF Labs device uses a full-custom ASIC chip originally designed as a IPSec VPN accelerator.

The intellectual property of BFL consists of finding how to internally modify the behavior of the chip
to change the IP protocol 51 (AH - Authentication Header) HMAC-SHA-256 to a simple two nested SHA-256 hashes.

Somebody please quote my message.

Thanks,

If it's too good to be true... :rolleyes:

I don't care about daily updates.  I care that they are working on their products.  They said the pre-order money would not be held in escrow but be used to develop the product.  I don't want daily updates, I want a product. Keep working BFL! 

No news from BFL yet again. 25 Nov I await anxiously !

It must be the second coming of Jesus but only a month earlier than usual .

God help us !

No, too Mediterranean and Westernized. Not the Bitcoin peace pipe either. I think nowadays offering of a bottle of beer is a thing that crosses all cultures.

For those who may be interested how people from different cultures react to the perceived slight on an interned board: have a look at the OK-Pay thread:

https://bitcointalksearch.org/topic/m.621678

and see how the response of a Russian (Eastern European) is so different from an response expected from a Westerner or Far Easterner.

A Bitcoin olive branch?

Thank you again for your valuable input. It seems to me that I have inadvertently touched some nerve of yours. I apologize for that. I'll definitely follow your advice, but not this year.

Our whole conversation here reminded me of the important lesson about the difficulty of the trans-continental personal relations. What is acceptable between acquaintances over a beer in a Western world is not acceptable even in an informal conversation in the Eastern world, especially over the long-distance communication links.

What this world needs more urgently than bitcoin is a bitbeer: a way to instantaneously transfer a small token peace offering over the trans-oceanic distances.

BTW, this company (if it even is one) claims to be working on an asic for bitcoin:
http://bitcoin.stackexchange.com/users/275/ken-simpson

Granted, I wouldnt hold my breath as any 15 year old could have made a more convincing website, but at least they arent taking preorders, so who knows . The CEO seems to be this guy:
http://ca.linkedin.com/in/ksimpson

Also founder and CEO of this company:
http://www.mailchannels.com/company/index.html

It may not be wallmart, but doesnt strike me as something run by a 15 year old punk or a nigerian scammer either.

We will have to see if anything pans out of it, but, like I said, Id think twice before plunging a boatload of money on FPGAs.

I dont expect to make money on my gpu mining. If I do, it will only be because of appreciation of bitcoin. Thats why Im not interested in expanding, what I do is mostly for fun and some speculation.

As for why buy FPGA's? I certainly wouldnt. Or maybe one for fun. But I dont believe they would pay for themselves before sasics arrive, since I suspect that might be really soon. What I can tell its really not extremely expensive, NRE is on the orders of 10s of $1000s. Like I said, if its not BFL, someone else will.

Would I jump on that? No, because I dont have enough faith in bitcoin's longterm sustainability. If I had, yes, I would buy that, and take the risk I get crushed by asics one day and pray I will have paid off those s-asics by then.

BTW, I wouldnt worry too much about moore's law. State of the art processes are prohibitively expensive, and increasingly so. If you havent earned back your money before the next process or two becomes affordable, you probably never will make money on it anyway.

Sure at astronomical risk.  Still for the sake of argument I will say yes ASIC will 100% definetly happen and render everything else obsolete. 

Given that why are you mining today?
Why did you even buy your first GPU?  You had to know that someday it would be obsolete.
Why are you interested in BF $500 board? (lets pretend it is real)

Both your GPU and BF board will be eventually rendered completely obsolete by future technology.
So why no sell everything and just wait for Custom ASIC to arrive?



Someday. However there is time to make money everyday between now and someday.  I mean you must already know this otherwise:
a) you wouldn't have bought GPUs.
b) you wouldn't be interested in FPGAs.

Wouldnt bet on it. Its expensive, but probably not out of reach to do small runs through MOSIS or CMP sharing masksets particularly using older processes like 90 or 130nm.


Have a look, from Altera's website:



Considering for bitcoin you barely need any IO or RAM, the efficiency increase would likely be much more than double. Looks more like ~5x to me. Depending on leakage perhaps more.


At 130nm....
If bitcoin becomes an actual currency, you could spend a fair amount developing a money printing machine.

Hardcopy electrical efficiency wouldn't be that high & full ASIC are not happening any time in the near future (if ever).  The greater threat comes from non-viability of Bitcoin (which would affect all custom solutions equally) and the relentless drive of Moore's law.  Still the comparison to rack full of Xeons isn't realistic.  The 7800 series cards will likely have 2x the electrical efficiency of 5000 and 6000 series cards.  The day 7800s launch does that suddenly turn all current rigs into obsolete unprofitable junk?  Hardly.  Sure a rack full of 7870s would be nicer than 5970s but the mining maket isn't that efficient to instantly make older tech obsolete.

Greater electrical efficiency starts becoming a game of diminishing returns.
My rigs get about 2.6MH/W and that is something I am pretty proud of.
Current real FPGA solutions are roughly 22MH/W.
Structured Asic is ~ double efficiency per watt say 40MH/W.
Lets split the difference on that chip and say 230MH/W.

1 Bitcoin electrical cost (above performance per watt & 1.2M difficulty)
GPU  - $0.88
FPGA - $0.10
SASIC - $0.06
Custom ASIC - $0.01

Sure SASICS and ASICS are more efficient however look at current prices they trade at roughly 3x my electrical cost and almost 25x FPGA costs.  Some day when a Bitcoin is priced at 3x SASIC electrical cost an FPGA is still profitable.   Eventually the combination of Moore's law and SASIC (or just more efficient FPGA) will drive the revenue below cost of production on a "current gen" FPGA cluster but that day is likely at least a decade away.  Remember even when the tech exists it will take time before the market adopts it.

TLDR version:
Scale doesn't negatively affect ROI%.
If a $20K FPGA cluster is going to lose money over its lifetime then ... a $500 FPGA board will also lose money over its lifetime.  So if you believe SASICS or Custom ASICS make building a cluster a bad investment why would you want to buy an $500 FPGA board?

Before you splash that kind of money on expensive and (relatively) power hungry FPGAs, do consider that if bitcoin is here to stay, sooner or later someone will do a (edit) easypath/hardcopy port, or even a  full custom asic. If its not BFL, someone else will, and it will make your off the shelve fpga's look almost as silly as  someone who bought racks full of xeons for bitcoin mining a year ago.

To get a feel for what full custom asics could achieve, have a look here:
http://rijndael.ece.vt.edu/sha3/chip/sha3-asic-datasheet.pdf

If my math isnt off, that test chip gets either 150 or 300 MH/W on an old 130nm process (1.51 Gbps SHA256 for 5mW). On 90nm you might expect double that, on 65nm you might get 4x that. 40 and 28nm, well, do the math, but those are probably too expensive for years to come.

Nice link.  Yeah something like that but with less specialized hardware.  Those boards provide 512MB of RAM per FGPA, DDR expansion, SDHC slot for local storage, high speed busses between FPGA and across backplane, low latency links between servers to build a unified cluster, built in server grade host PC, expansion slots, high speed connectivity to SAN and LAN, etc.

All that adds a lot of cost and is totally useless for Bitcoin.

So take a design like that and make it "ghetto", use a low speed serial to connect the FPGA across the backplane to a USB micro controller, drop everything that isn't necessary and include no host PC.  Just a chassis full of FPGA connected to backplanes.  Then connect power to each backplane and run a single USB cable from each backplane to a host PC (in a seperate chassis).  To expand just keep adding modules until the chassis is full, and then add another chassis.  A single host PC could connect to an entire rack of FPGA chassis.

Just to remind our "modular FPGA miner" project just followed this approach(see the link in my postunderline).
But it seems there wasn't enough interest. We got stuck summer this year.
Im still pursuing it, but alone im not that fast you know

I quoted the copacobana system multiple times.
It was the precessor to the system wich are now produced by SCI www.sciengines.de
I got qoute for that system starting at 12000€.
It is using spartan 6 FPGA's said to deliver 200Mh/s each and 6 FPGa's per board.

so its not that new

Scalability.  My "end game" would be putting 4GH/s in a 4U chassis and dropping it in a datacenter.  There it is protected from theft or loss, has reliable electrical power & connectivity.  Also easier to get an insurance policy on a $20K cluster if it is in a protected facility. It would just sit there hashing until the end of time.

Larger cards make that a little easier to achieve than hundreds of small boards.  I agree if the goal is to buy 200MH or maybe even 1GH small cards are "easy", but what about 10GH, 20GH, 50GH?


Would using a backplane work "better"? That would allow better economies of scale and each board would be "cheaper" however you could still get some real densities.

Something like..
http://www.copacobana.org/




Granted it wouldn't be exactly the same.  We are talking about larger FPGAs with higher thermal load so some heatsinks would be required and that means greater distance between cards to allow proper airflow.

Still the general concept would be putting a backplane for multiple cards in a 19" 3U or 4U chassis.  Maybe 18" backplane with 5 cards (10 FPGA) spaced 3" apart.  Then mount two backplanes per 4U chassis (perpendicular).  That would be 20 FPGA.  For LX150 we are talking 160W and 3.5 to 4GH/s.

Not to take this scam thread totally off topic but how much more expensive would it be to adapt existing board so that it can get connectivity (maybe serial/COM ) and power from a backplane.  The backplane then could simply have a PCIex16 connector, a USB controller and communicate to the 10 FPGA via serial bus.  Cooling could be acheived by using 5x 90mm fans at the front and back of the chassis (maybe a third "row") between the two FPGA backplanes.  By moving the USB to backplane that would simplify board design right? Would it be possible to move power supply to the backplane too making each board even simpler.

Am I crazy?  Would it be too much cost/complexity for little gain.  I think it might even be possible to get 3x FPGA per board putting density at 5 to 6GH/s.  If you don't want to derail this thread feel free to drop me a PM.

Likely because they have a significant NRE and a small per chip cost. If they use something like Easypath or hardcopy, their per chip prices could be up to 80% lower than off the shelve FPGAs, but they would have to aim for many thousands of boards to recover the NRA.