Topic: BFL - Update 3/4 (Read 5053 times)

While this is undeniably true, things which change your financial projections are a big deal and can have massive implications for the future of your company.  While reducing the number of chips per device doesn't significantly reduce the cost per unit of those devices, it does allow you to build significantly more devices at $1300 each (or to fill more orders which have already been paid for).  If they can scavenge 16,000 chips out of the 74,000 by reducing the number of chips per unit, that's 2,000 more Single orders which can be filled without having to order additional chips - that represents an $2,600,000 in revenue for the first production run, which is significant by anyone's criteria.

While BFL is a bit of a black box in terms of information regarding its funding sources, nothing they have said suggests that they have access to a bottomless well of cost-free money.  One way or another, these delays are costing them significant money and they have to find that money from somewhere - whether it's from pre-orders, selling equity, or loans.  Whatever money they're spending now on unanticipated costs is money which won't be available for something else in the future.

I agree that Josh has appeared not to know a lot of material information until the last moment, but it's been obvious for months that Josh is not directly over-seeing this project and that he gets much of his information second or third hand.  Josh being unaware of something doesn't mean that those actually involved in the step concerned had incomplete or inaccurate information (hell, Josh seemed surprised to learn that piddly little orders aren't going to be given priority when it comes to facilities allocating production time at short notice).

BFL customers have been somewhat over-involved during the last 9 months.  How many chips were being produced/had been ordered is not information which would normally be shared with customers, especially when no information has been given about how many of each unit type are going to be produced from the first 5,000 chips and the following 6,000.  People have reacted to this over-sharing by feeling entitled to have a voice in the business decisions made by BFL.

Given how little we know about BFL's process and history, it's hard to say anything with any certainty. We know that BFL is only paying their contract manufacturer on completion of a successful project, so it could be just as likely that BFL is paying a fixed cost per chip as opposed to per wafer. Hell, for all we know there might not even have ever been 100k chips. They didn't know their first batch of wafers off the line was 6 instead of 12 until they were almost in the air.

Regardless the cost of the actual chips in a Single is going to be a pretty small part of the $1300 they charge for it. While having to eat whatever the cost of the wafers was will hurt their bottom line, going from 8 chips to 7 chips is not going to be a big cost savings.

They can fill less orders with 74,000 chips than with 100,000 chips, so their financial projections and their break-even point are going to be significantly affected by the diminished "first run" revenue.  A 25% reduction is revenue is huge and it needs to be offset as much as possible as it's very likely some of that revenue was earmarked for meeting costs associated with the second mass production run. 

Ordering another 100 wafers requires having the money to pay for them (and the cost of packaging and assembling them).  We simply don't know whether any funds from existing pre-orders have been spent over the last 8 months or the amount of revenue from first run sales which was earmarked for the second production run.

If the QFN to FCBGA conversion goes well and BFL doesn't need to redo the design, they really only need to worry about the quantity of chips they have on the way if they are going to be able to ship and sell all 75k before they can get another set of chips done. With the design being done there's no real reason they couldn't get started on another 100 wafers and have finished chips delivered in June.

They haven't just sacrificed one wafer, though.  The combined impact of various decisions made by BFL since late last year means that instead of Batch 1 plus their bulk order yielding 100,000 chips, Batch 1 plus their bulk order will now yield 74,000 chips. 

You cannot make the same number of units with 74,000 chips as you can make with 100,000 chips unless you use less chips per unit - and the response to Josh mentioning that possibility suggests that BFL will lose orders if they reduce the amount of chips per unit.  It's hard to know how many people would follow through and actually cancel their orders as many customers have previously stated that they'll bail "if X occurs" and then not done so, but it's very clear that a non-trivial amount of customers planned to over-clock the fuck out of their units and are deeply unhappy about that possibility being removed or diminished.

new pics
https://forums.butterflylabs.com/dbtgallery.php?do=browse_user&userid=1&gal=gallery&id=7&type=albums

Yep.

I think it is the first time I've hit the ignore button on someone with such a low post count. 

I'll not be put in the position of defending BFL. Let em have it, they're completely full of sh1t.

You lose credibility immediately however when you lump jgarzik, friedcat, eleuthria, and bitcoin foundation in with the butterfly cowboys.

It's funny . That i am newbie in this forum does nothing to what i am and what i can . It is not right for me as a company onwer to say that but as far as i know BFL sales  mining equipment since 2011 and they even dont have any prototype yet, and they will ship thausends of them in any moment, very funny. There are some pictures of ASIC miners  like  PCB's and boxes, but no VIDEO that shows  it actualy working! If i have build working one l would make many  videos that shows how it work realtime with statistics, yeah, definetily would  do that like all the good guys in you tube that build bitcoin farms with many GPU's.
for me 67117 is proxy pool with big rejection rate.
It is fine to call me newbie. 

https://bitcointalksearch.org/topic/m.1572769

TAKE MY MONEY NOW!

Umm...since you essentially called friedcat and eleuthria liars and they have just a wee bit more street cred around here than a newb using an account created last week...I'm gonna have to go ahead and ask you for some kind of evidence on this. Nothing personal you understand.

This:

https://bitcointalksearch.org/topic/asicminer-user-ranking-on-btc-guild-148162

...btw sir sockpuppet does NOT count as evidence.

110nm is an old  technology. You cant expect GH from that. If it was possible the miners with Cyclone and Spartan FPGA already can be made. They aren't so dont be fool to be involved in this ASIC hoax. Also 67117 in btcguild is proxy pool.
regards.

BFL is using a 300mm wafer.

BFL's die is 7.5mmx7.5mm.

I also suspect the BFL package is 15x15, not the die. 15x15 would be quite large and the die seems rectangular. 15mm is a standard package size though.

That is one damn tiny chip!

(If it is ASICMiners specs...wow, they must have a hell of alot of chips worth to bring online.)

I believe Avalon's die size is 4x4 and the packaged QFN size is 7x7, but I don't feel like looking it up. I might have ASICMiner's die size instead, but either way the actual die size will be less than 7x7.

I don't recall. I think it may have been 4,000 chips per wafer. I only recall that the die size is 7x7mm while BFL's is 15x15.

Sounds like they buffered the clocks too much