BFL 3-15-13 update...Testing to Shipping...in "days".

[smoothie]

https://forums.butterflylabs.com/announcements/692-bfl-asic-status-2.html

It's been a long couple days here at the labs! Trying to test the chip in the test rig has been exceptionally trying and it turns out it's due to a bad socket on the tester. Initially we were concerned there might be a problem with the bumping or substrate, but fortunately we had tested the chips prior with the wire bonding technique, so we knew they worked. Getting the test boards made proved to be fortuitous as well, since we were able to bypass the test rig completely and bring the chips up on the board. By doing so, we were able to finally identify that the problem was related directly to the test rig and not any intermediate process or step.

Bringing a chip to life in situ on a board is not the easiest thing, so it has been slow going. The chip was responding properly on the board late this afternoon and we will be picking up the process in the morning. We hope to have a more complete test by Saturday night or Sunday sometime. Meanwhile, we will also be hard soldering a chip into place for use in the test rig and replacing the socket to allow the bulk testing to finish. In some other positive news, we've not found a single bad chip yet, which could mean our yield rate will be exceptionally high... maybe we just got lucky out of the 50 chips we have available on boards so far, but it seems unlikely. So that may mean the vast majority of our chips will be usable.

So the good news is the boards work, the chips work, the bumping works, the substrate works. We just need to nail down a bit more with the firmware and we should be able to conduct a full test and start shipping.

I will be posting another update as soon as I have more information. Needless to say it's been a pretty busy and stressful last few days here, and some script kiddie deciding to DDoS our sites this morning didn't help matters. All connectivity problems should be resolved now and we will keep people updated as time allows.

I find it odd that BFL is going to go from testing to shipping in a matter of days (from the sound of it like a "few" days). This is where Josh keeps failing.

They dont even have a working prototype and it is MARCH......TWO THOUSAND AND THIRTEEN. Back in November they were leading on making it seem they had a prototype then but couldn't ship until 30 days or more. So how is it that all of a sudden that they FINALLY get a "working" prototype that they are going to go from TESTING to SHIPPING in a few days?

Expect more delays people. Everything BFL made themselves out to be has come out to be false. They outsourced pretty much every part of the ASIC development process to a third party. And that is just to get started...

BFL is and will continue to be full of crap as long as they DONT disclose what actually went on during the time frame of October, November, December...

So fail...

[Bitobsessed]

Oh, they will ship.  Just to themselves...in...god knows where.

[lolwut]

So they have a board and a chip that they claim works, but they dont have a 100% working firmware, nor have they *actually tested* the chips with the firmware and drives, yet they claim the chips work. Its kind of hard to claim the board and chip work if your firmware is buggy and you haven't conducted a full test. They act as if testing will go quick and smooth, this is NEVER THE CASE.

CAVEAT EMPTOR

I'm sure they are working on a product, and I'm sure they even have some good engineering samples, but from their updates, theres no way they can possibly ship anytime soon unless they cut a lot of corners and release a half-baked product.

[Frizz23]

... and release a half-baked product.

I wouldn't wanna be a batch #1 or batch #2 customer aka BFL beta tester.

[witherworth]

... and release a half-baked product.

I wouldn't wanna be a batch #1 or batch #2 customer aka BFL beta tester.

It sounds like most of the issues are in the firmware, which means that even if the first units have to run at 30GH/s, when the firmware is fixed, we'll all be running at 60+. Software is a lot easier to fix than hardware.

[bushstar]

This update is not surprising. BFL has been way too optimistic every step of the way. I look forward to them shipping to see an end to the constant steam of conspiracy theories and scam accusations.

[Aseras]

"days" is like Soon™ right?

[Red_Evil]

All the technical details are wonderful but as a customer all I really need is a ship date for confirming product.....

[witherworth]

They either need to ship in the next week, or they need to ship in 3 weeks, after I've moved. It'll be my luck that I'll move, try to update shipping address, only to find out that the units have already shipped to the old address, then I'll spend weeks trying to recover my packages... More waiting I suppose. I should be pretty good at it considering I'm still here from a 2nd month preorder.

[Raoul Duke]

They either need to ship in the next week, or they need to ship in 3 weeks, after I've moved. It'll be my luck that I'll move, try to update shipping address, only to find out that the units have already shipped to the old address, then I'll spend weeks trying to recover my packages... More waiting I suppose. I should be pretty good at it considering I'm still here from a 2nd month preorder.

Tell the postal service to forward everything to your new address.

[Logik]

Has anyone told BFL that they aren't actually selling ASIC chips by themselves? That's what it sounds like they think they are doing "The chips work".. well, cool, but you have to like, ya know, actually place them on a board and then run software on it and package it etc. Minor details I guess.

[meowmeowbrowncow]

So they have a board and a chip that they claim works, but they dont have a 100% working firmware, nor have they *actually tested* the chips with the firmware and drives, yet they claim the chips work. Its kind of hard to claim the board and chip work if your firmware is buggy and you haven't conducted a full test. They act as if testing will go quick and smooth, this is NEVER THE CASE.

CAVEAT EMPTOR

I'm sure they are working on a product, and I'm sure they even have some good engineering samples, but from their updates, theres no way they can possibly ship anytime soon unless they cut a lot of corners and release a half-baked product.

And, for god's sake, let the LED's work or lolwut will lose his shit.


(lolwut likes to back out on thousands of dollar deals when flashing lights don't work.  I guess it's a 'oooo, shiny,' kind of response.)

[Beta-coiner1]

"days" is like Soon™ right?

I don't know why but I got the impression for moment BFL was the Duke Nukem of ASICs.

[PuertoLibre]

Sir, At least Duke Nukem Forever shipped...ten years later.

Who knows if BFL ever will.

I feel you have soiled the Good name of the Duke. The resemblance is only in passing.

https://www.youtube.com/watch?v=MzxWZupHWkc

(Warning, there is a lot of Joshin' in the video.)

(Bonus: https://www.youtube.com/watch?v=UObpMvgmHjM)

[Phinnaeus Gage]

Sir, At least Duke Nukem Forever shipped...ten years later.

Who knows if BFL ever will.

I feel you have soiled the Good name of the Duke. The resemblance is only in passing.

https://www.youtube.com/watch?v=MzxWZupHWkc

(Warning, there is a lot of Joshin' in the video.)

(Bonus: https://www.youtube.com/watch?v=UObpMvgmHjM)

Gosh, Wally, I thought you were Joshin'.

[Bicknellski]

Bitcoin Mining Rig
- www.butterflylabs.com/ - Blocked
We will not show you ads from butterflylabs.com again.

[johnyj]

After seeing Avalon's gigantic power supply design and huge heatsinks, I start to worry that the BFL products could not be easily run at advertised frequency (Maybe jalapeno and small single can do)

I remember that someone discussed before, going from a 110nm process to 65nm process in principle won't generate a 80% increase in efficiency, the best estimate is around 50%. There was a possibility that Avalon had an unefficient design, then BFL might gain another 30% efficiency by using a better design

But now I have closely examined Avalon product, I think their design is already very modulized and efficient (although not very delicate, but carefully planned to reduce the loss of power), BFL won't gain too much efficiency by a better design

Take these numbers into count, each BFL chip roughly equals to 1 avalon small unit (10 chips), containing 16 of much smaller sized engine, and that single chip will require similar amount of power supply that 1 avalon small unit requires (about 25W), providing 50% more hash than avalon small unit (2.75GH), e.g. around 4 Ghash

small single 4x25 W = 100w
single 8x25 W =200w

Suppose these can be further trimmed down due to 65nm's lower heat generation, let's say that 120W would be needed for a single, then the current power supply design on the board will not be sufficient, if you run it at 60W, then the hash rate will drop by half

So I guess BFL might face a decision of changing the power supply design (which will cause months of delay) or lower the claimed performance and get customer some refund/discount, or any other solutions in the middle

[mrb]

I remember that someone discussed before, going from a 110nm process to 65nm process in principle won't generate a 80% increase in efficiency, the best estimate is around 50%. There was a possibility that Avalon had an unefficient design, then BFL might gain another 30% efficiency by using a better design

In principle going from 110nm to 65nm reduces power consumption by 65% (1-(65/110)**2  because power efficiency is proportional to the inverse of the square of the feature size).

But the process node is not everything. Consider ASICMINER for example: they have chips that are more power efficient than Avalon, despite the worse process node (respectively 130nm vs 110nm). ASICMINER does 167 Mhash/Joule vs Avalon doing only 150 Mhash/Joule.

[creativex]

I wonder if ASICMiner gains an efficiency advantage here by not shipping end user gear. It will be interesting to see if they can maintain those numbers when they do.

[johnyj]

I remember that someone discussed before, going from a 110nm process to 65nm process in principle won't generate a 80% increase in efficiency, the best estimate is around 50%. There was a possibility that Avalon had an unefficient design, then BFL might gain another 30% efficiency by using a better design

In principle going from 110nm to 65nm reduces power consumption by 65% (1-(65/110)**2  because power efficiency is proportional to the inverse of the square of the feature size).

But the process node is not everything. Consider ASICMINER for example: they have chips that are more power efficient than Avalon, despite the worse process node (respectively 130nm vs 110nm). ASICMINER does 167 Mhash/Joule vs Avalon doing only 150 Mhash/Joule.

Check this:

However, I must warn you (Inaba), don't place trust in the wrong people and what they tell you; for when that moment comes you won't able to meet the advertised specifications, at no real fault of you own besides ignorance since you really believed BFL. Be ready to take the spear to the chest as you were their spoke person. I wish you the best of luck and hope BFL can ship on time for when that time comes this forum can go back into the ol' days of competitive spirit of reducing power consumption and increasing hashrate of each others product.