Bitcoin Forum

Edit 7-Nov-2012: BFL_Josh has posted both figures, so this bounty is now closed.

I will offer a 20BTC bounty to anybody who posts a measurement of the size of the die used in BFL's "SC" devices.  I'm assuming they all use the same chip.  It's really easy to get the die measurement from a broken device -- just grind off the plastic/ceramic top using a dremel and pull out a ruler.  I will be happy to escrow the bounty with a well-known forum member if you can present credible evidence that you're serious about this -- that way you can be sure you'll get paid before you totally wreck your already-broken device.

I need this figure in order to compute the η-factor (https://bitcointalk.org/index.php?topic=119668.0) for their product.

I will also offer a 20BTC bounty for knowledge of which process node BFL is using.  However, I caution in advance that it's not as easy to be 100% certain about that -- you can't just measure it with a ruler -- so claiming this bounty will be more difficult and will involve more verification (which might not even be possible) than the raw die-size bounty.  Please post here if you have any questions about what would be sufficient to claim this bounty; I don't want any hard feelings.

I have offered bounties like this in the past and have a history of paying promptly (https://bitcointalk.org/index.php?topic=79825.msg886359#msg886359).

Both of these bounties become null and void if BFL releases the figures themselves.

so you expect someone to get the die size from pics alone as the SC isn't out yet?

Obviously I am not expecting this bounty to be claimed in the next few weeks.

Also, from past experience with bounties like this, the sort of information I'm looking for tends to bubble to the surface in ways I never expected.  So it can't hurt to post the bounty early.

With a "lifetime warranty", why would someone chop up a unit just to give you your information, rather than sending it back to BFL for a new one?

Also, if you're giving out 20BTC for a bounty, why not rather just spend the 15BTC for a Jalapeno directly from them?

Probably because they broke it.

Warranties don't cover abuse.  For example: got it wet, left it in an enclosed space without a fan where it overheated, wrong power supply, etc.  Yes, these things are rare, but I only need one person to damage one device out of BFL's entire customer base.  And I'm sure that poor soul will be glad to get their money back (and then some).

Also, as I mentioned earlier, there are always possibilities I couldn't imagine.  Somebody mailed ngzhang a broken BFL-Altera device out of the blue, which is how we got the IDCODE off of the device back when they were so expensive people were afraid to go near them with a soldering iron.  Could happen again.

Alternatively -- and less likely -- maybe somebody else is curious too and has the tools to do this without destroying the chip.  Those are QFN packages which means you can remove the lid and put it back on, though it's not easy.



Probably the same reason why you aren't arbitraging the situation to pocket an easy 5 BTC.  ;)

I could order a jalapeno, but I wont because I don't have the cash. Seeing as you are offering a bounty, you do, so why don't you?

He already answered that.

no, he said "probably for the same reason you aren't"  (paraphrasing)
The reason I am not is because I don't have the money, but here he is offering a bounty for more than the price of a jalapeno, therefore he has to have the money to buy one, unless he is intending to scam. Or am I missing something?

I think he is hinting on an insider leaking info to him

Hi Elden, the node and method are detailed in the upcoming issue of bitcoin magazine.   (Issue #4).

Probably 90nm. Seems to be the most cost effective.

Prepare for the improbable.  :)

14nm? That'd be pretty improbable. ;)

Wouldn't that be nice...

I believe my prediction of 65nm (https://bitcointalk.org/index.php?topic=95762.0) will be shown to be right :)

It'd be funny if BFL won this bounty.

LOL Josh mails him a Jalapeno.

sorry dude, i invested with your competitor(who is using 90nm apparently)... will be interested to see what you've got 'under the hood' either way.

May the force be with you.

Maybe 45nm? That could be why Josh sees these as penultimate without room for much improvement.

28nm! I want bounty for random guessing!

I'll tell for 40 BTC!  Do I hear 50? :)

Smart thinking.
I believe you are right about this, after reading BFL posts.
45 is still too expensive, even AMD is stuck at 32.
Only Intel is capable to do 22 at the moment

Inaba, you cannot get paid for that, because...


It was very craftily formulated to not let you earn a single satoshi. ;)

If Inaba tells me the figures but not in public then I can collect.  I'll only charge 50%, how about it Inaba?

I'll undercut at 45% :D

Damn, my plans to collect the booty are foiled! 

45%... do I hear 40%?

Nice .....LOL

He is Inaba, not BFL_Josh.

Lol - Does this qualify as a "Booty call"?  :)

There is ALWAYS room for improvement, if not we would all still be living in caves eating rocks.

0nm

I win.

lol. good one.

I don't see a die-size prediction. I'm going with 24mm^2

From the board pictures, the chip packages appear to my eyes to be 11x11mm.

Not that that says a whole lot about the die size.

40nm
no 37
maybe 5?

Is the answer 42?

The answer is ALWAYS 42.

Perhaps -1 nm.

Or sqrt(-1) if you think BFL is creating imaginary ASICS...

I'm sticking with 24mm^2 at .065 micron.

A source tells me that the Bitcoin Magazine article alludes to a 45nm or smaller process.

Not very concrete.

I can't believe there hasn't been someone who's claimed this yet.

It's simple, those that know have an NDA.
(I've seen several of these for contract positions that have penalties that can actually add up to several times the wages paid for the contract if the NDA is violated. So non-disclosure gets kept (and the consultant gets paid a premium for keeping his mouth shut.))

Until official release we are unlikely to get confirmation.

The information is apparently in Bitcoin Magazine, and subscribers have reported getting the magazine already. A quick readthrough and post is worth 40BTC.

Or they are honorable people who value the trust and respect gained from not breaching confidentiality for some minor prize even absent a formal contract.

Potentially true, but since they have investors I'm pretty sure they would insist on having any commitments like that spelled out contractually. NDA's in particular have a psychological impact that helps folks remember to keep their mouths shut, and the assurance that the other actors in the company (coworkers) have the same motivation, without having to know them personally and trust them.

Even in a small organization, it is a good idea to spell out the rights and duties and often a general code of conduct just to make sure that everyone really understands. It also protects you from sometimes down the line when you have a falling out with someone in the organization, or a partner dies and their heir is an idiot.

I'm not saying you are wrong at all, just that given the past (Sonny, Costa Rica, etc.) and everything I have seen (except Inaba sometimes) leads me to believe that they have a pretty good understanding of the power of contracts and are used to using them.

It's like "Trust but Verify" principle often used with teenagers, except when you have a written contract you don't have to rely on the troll-patrol to call you out, you go to court. Not everyone is honorable, or at least not all the time under every circumstance, so a little insurance is a good thing for the whole company.

So, who has received the magazine?

Psychoticboy has said his arrived.

which issue? 4?
I might buy it, maybe. I dunno.

Sounds like BTCCurious got one, too.
https://bitcointalk.org/index.php?topic=83530.msg1311316#msg1311316

It's lame that we have to pay for information that BFL should tell us for free....

Or I guess I could be patient and just wait for it to come out. But, where's the fun in that?

I got my copy yesterday! hahahahahahaha!
I'm seriously impressed with the bitcoin mag. Once my ASIC arrives I know where I'll be spending some BTC.

As for the die size I just wanna say good choice! Efficient but room to grow.

So have you claimed the bounty?

I thought the bounty was nullified when the Magazine started hitting. See the last line:


BFL released the specs to Bitcoin Magazine who got to do an "Exclusive" on it. Now all the other Bitcoin Magazines can report on it too!

Nope.
 The Magazine is really good and I would recommend buying it for more than getting a few more stats on BFL hardware. Good price point just about any small time miner like myself can afford it. It is a good way to help grow the btc economy.

We would appreciate it if you would share the basic info, having it locked up in a dead-tree publication is stupid. We have little enough happening that I'd rather not see some huge game being played with a few people hoarding information just to act like trolls.

If they sell the magazine electronically (even for fiat) I'm there, but I get too much dead-tree as it is.

FYI, there is no moratorium on the info re: Josh.

Do you understand the reason for this thread and the reasons we're hoping you posted in this thread?  Please enlighten us on the BFL article.  Specifically the SC die size and/or process node. 
Gracias.

A user on the BFL forums posted the entire article and it doesn't appear to contain any information about die size or process node. The only interesting thing it says is that BFL is considering a future, next generation ASIC offering that moves to a smaller 45nm process node, or possibly even smaller than that. So we know that the process node they're using is bigger than 45nm - which we could've guessed anyway - but that's it.

I'm going to guess BFL is using 65nm, seems to be a sweet spot between chip price, performance, and power consumption. Not that I really care though, I went with a manufacturer that's more transparent throughout the whole process(bASIC/cablepair/Tom).

http://bitcoinmagazine.net/bfl-confirms-65nm-process-for-sc-lineup/

No, because the bounty clearly states (and has stated since the very first day it was posted) that it is null and void if BFL releases the figures.  BFL has released the process node, which is 65nm, so that bounty is no longer valid.

The die size bounty is still open.

Didn't Josh say 3cm x 3cm, or was that external chip size, not die size?

Link?

I thought the package size was 11mmx11mm.

30x30 is huge.

Yes, sorry, 11mm x 11mm.  My memory was thinking of something else.

https://bitcointalk.org/index.php?topic=119351.msg1311322#msg1311322

The die size is 7.5mm^2

Where's the source for that then?

Tell you guys what.  I can confirm process node and die size if I can get a chip.
I can get the die size with XRD and process node with cross section SEM.
Anything thing else?  # of Metals, bonding wire method, packaging durability,
well profile, die thickness

Check the comments section.  BFL_Josh clears up the die size question. 
https://forums.butterflylabs.com/content.php/125-BFL-ASIC-Update

The die size is 7.5mm^2

Hrm, that would be a square die less than 3mm on each side, which is pretty hard to wirebond and in a package that would be ridiculously oversized for it.  There are sometimes power and thermal reasons for using an absurdly large package, but putting a 7.5(mm^2) die in an 11mmx11mm package means you're going to have bonding wires on the order of 4mm long on all sides -- at that point you have some serious inductance in the bond wires.  Not saying it's impossible, but it certainly is unusual.  That's the very upper limit (http://www.mosis.com/pages/products/assembly/index) of what MOSIS allows.

Can we get BFL_Josh to confirm that he didn't mean (7.5mm)^2 aka 56.25(mm^2)?  He's transposed those two units before claiming the package was 11mm^2.

7.5(mm^2) isn't impossible, but 56.25(mm^2) is more plausible.

He confirmed it on the BFL thread referenced earlier. 56.25mm². 7.5x7.5mm.

Thank you, that makes way more sense.  Somebody please teach that man the difference between mm and mm².