ASICMiner BE300S Samples Arrived, <0.2W/G Achieved at Board Level

[dogie]

does anyone know if asicminer has 'taped out' and produced masks for the be300?  ie, is it in production?

Last info was that first batch of BE300 will be done in February. They must be in production now.
https://bitcointalk.org/index.php?topic=833704.msg9678661#msg9678661

thanks... its not quite the same thing.  do we know who amhash is?  do they represent friedcat or are they a customer?   all we know is that someone (who wasn't friedcat) said in november they would have chips back in feb.  but until we hear it from the horses mouth or from someone definitely representing him, then i don't think we know for certain that they've taped out and will definitely have the chips back in feb.  or whenever.   all we know is they continue to test the samples, but the samples aren't from a production batch and are independent from that.

 

If you read relevant threads, you'd see answers to your questions.

thats cryptic and not exactly helpful.  have you seen a tapeout announcement or friedcat recently stating availability of production chips?

Tldr AMHash = ASICMiner

[aerobatic]

Tldr AMHash = ASICMiner

anyway, my point is that unless you hear it from friedcat that they've taped out, why would you assume they've taped out yet?  no where does it say it.   and even in the post from amhash in nov that says silicon due in feb.. that was presumably before tapeout had occurred so theres a lot of uncertainty in that statement and its datapoint is old and stale.  it costs a lot of cash to tapeout, and with bitcoin at the price it is now, the economics to justify taping out are stacked against you.  sure you could do it, but would you spend millions pushing that button in the current climate?

[Mabsark]

Update

Development
The 28nm BE300 engineering batch tapeout was in September 16 with TSMC.
...
14nm/16nm projects are at pre-evaluation and pre-design stage. It has to be done
in 2015 but the starting time of placing orders depends on overall gain vs (NRE/R&D/risks).
...
The mass production time of BE300 in terms of chip-out date is February to March, 2015.

December 16 is to be expected for us to get the chips. Testing time varies at 3-10 days since we had much more preparation work already done this time.

They got those chips about a week early it seems so they seem to be on track and we've heard nothing that would suggest any delays.

[aerobatic]

Update

Development
The 28nm BE300 engineering batch tapeout was in September 16 with TSMC.
...
14nm/16nm projects are at pre-evaluation and pre-design stage. It has to be done
in 2015 but the starting time of placing orders depends on overall gain vs (NRE/R&D/risks).
...
The mass production time of BE300 in terms of chip-out date is February to March, 2015.

December 16 is to be expected for us to get the chips. Testing time varies at 3-10 days since we had much more preparation work already done this time.

They got those chips about a week early it seems so they seem to be on track and we've heard nothing that would suggest any delays.

those are engineering samples.  they come from a set of wafers called a Shuttle - or MPW - Multi Project Wafer.  these are wafers that contains lots of customers samples on the same wafer.

They're not production wafers, and the huge NRE will not have had to be paid for those samples.

For the real production, a multi million dollar NRE will have to be paid, plus an order for several 'lots' of wafer production (also multi million dollar).   its a big commitment at a time when the price of bitcoin is very low and the economics of taping out a new 28nm chip are a painful decision.  It'd be surprising if friedcat has pressed the button on this expensive tape-out and production, especially when he knows that bitfury has a better chip (and presumably others) also out in the same timeframe.

[Mabsark]

Like I said, full production was scheduled for Feb-March and samples were early. There's been no indications of any delays so why would you assume that the schedule has changed?

Do you have any evidence to show that BitFury has a better chip that will be out in the same time frame? BitFury (and others) said that they'd have 0.2 J/Gh ASICs (based on simulations) in a similar time frame. Have you seen any evidence of that? We know that AM have had 0.2 J/Gh chips since Dec 10.

[aerobatic]

Like I said, full production was scheduled for Feb-March and samples were early. There's been no indications of any delays so why would you assume that the schedule has changed?

Do you have any evidence to show that BitFury has a better chip that will be out in the same time frame? BitFury (and others) said that they'd have 0.2 J/Gh ASICs (based on simulations) in a similar time frame. Have you seen any evidence of that? We know that AM have had 0.2 J/Gh chips since Dec 10.

i have no evidence, only hearsay.

we do know, as you say, that AM has samples back that are 0.2 j/gh... but as mentioned, we don't know when they taped out (nor if they have), and production will be approx 2 months after they tapeout and pay the NRE.  they could still in theory deliver march, just, if they tapeout this month.  but unless they've already taped out in december, the chance of them having production silicon in feb isn't too good.  and now its march you say, someone else was quoting 'amhash' as saying it was feb.  both dates are possible if they've taped out.  so do we know if they've taped out?

[Mabsark]

The AMHash quote is from a later date than FriedCat's so it could be based on more recent information. Like I said, there's no reason to assume that the schedule has changed from a chip-out date of Feb-March. They got the chips on Dec 10 and testing was supposed to take 3-10 days.

[philipma1957]

Asic Miner will be late not BFL late but late as in 2 to 5 weeks.

Zero evidence on my part but they have had a pattern of a little late more then once.
So I figure history will repeat.

Hoping this gear does not over heat and run crazy hot. I want to have one in my hands sooner rather then later.

[Rabinovitch]

So as most of us I guess. 

[iCEBREAKER]

Like I said, full production was scheduled for Feb-March and samples were early. There's been no indications of any delays so why would you assume that the schedule has changed?

Do you have any evidence to show that BitFury has a better chip that will be out in the same time frame? BitFury (and others) said that they'd have 0.2 J/Gh ASICs (based on simulations) in a similar time frame. Have you seen any evidence of that? We know that AM have had 0.2 J/Gh chips since Dec 10.

i have no evidence, only hearsay.

we do know, as you say, that AM has samples back that are 0.2 j/gh... but as mentioned, we don't know when they taped out (nor if they have), and production will be approx 2 months after they tapeout and pay the NRE.  they could still in theory deliver march, just, if they tapeout this month.  but unless they've already taped out in december, the chance of them having production silicon in feb isn't too good.  and now its march you say, someone else was quoting 'amhash' as saying it was feb.  both dates are possible if they've taped out.  so do we know if they've taped out?

How can you have sample chips without a tape out?

[aerobatic]

Like I said, full production was scheduled for Feb-March and samples were early. There's been no indications of any delays so why would you assume that the schedule has changed?

Do you have any evidence to show that BitFury has a better chip that will be out in the same time frame? BitFury (and others) said that they'd have 0.2 J/Gh ASICs (based on simulations) in a similar time frame. Have you seen any evidence of that? We know that AM have had 0.2 J/Gh chips since Dec 10.

i have no evidence, only hearsay.

we do know, as you say, that AM has samples back that are 0.2 j/gh... but as mentioned, we don't know when they taped out (nor if they have), and production will be approx 2 months after they tapeout and pay the NRE.  they could still in theory deliver march, just, if they tapeout this month.  but unless they've already taped out in december, the chance of them having production silicon in feb isn't too good.  and now its march you say, someone else was quoting 'amhash' as saying it was feb.  both dates are possible if they've taped out.  so do we know if they've taped out?

How can you have sample chips without a tape out?

They have produced samples using a multi project wafer. They haven't had to pay  the high nre for the masks to be produced.

A shuttle run (mpw) costs only hundreds of thousands of dollars as the costs are shared with lots of other manufacturers. Each customer only gets a few samples of their chips.  Maybe a few hundred or less.

Once the samples are tested and everyone's happy with their chip. And no further changes are made then the nre is paid and full mask set is produced. That costs ballpark 2-3 million and the first order for a bunch of wafers probably also could cost 2-3 million so the initial order with nre and mask set could easily be a 5-6m cash payments (in advanc) for 28nm. Thus it's unknown at this time whether asicminer has taped out and paid the nre for the masks or for production wafers to be made. All this takes 2-3 months.

In bitcoinland many ASIC companies havnt bothered making a shuttle and have gone direct to production. Higher risk but it has usually worked. 

[iCEBREAKER]

Like I said, full production was scheduled for Feb-March and samples were early. There's been no indications of any delays so why would you assume that the schedule has changed?

Do you have any evidence to show that BitFury has a better chip that will be out in the same time frame? BitFury (and others) said that they'd have 0.2 J/Gh ASICs (based on simulations) in a similar time frame. Have you seen any evidence of that? We know that AM have had 0.2 J/Gh chips since Dec 10.

i have no evidence, only hearsay.

we do know, as you say, that AM has samples back that are 0.2 j/gh... but as mentioned, we don't know when they taped out (nor if they have), and production will be approx 2 months after they tapeout and pay the NRE.  they could still in theory deliver march, just, if they tapeout this month.  but unless they've already taped out in december, the chance of them having production silicon in feb isn't too good.  and now its march you say, someone else was quoting 'amhash' as saying it was feb.  both dates are possible if they've taped out.  so do we know if they've taped out?

How can you have sample chips without a tape out?

They have produced samples using a multi project wafer. They haven't had to pay  the high nre for the masks to be produced.

A shuttle run (mpw) costs only hundreds of thousands of dollars as the costs are shared with lots of other manufacturers. Each customer only gets a few samples of their chips.  Maybe a few hundred or less.

Once the samples are tested and everyone's happy with their chip. And no further changes are made then the nre is paid and full mask set is produced. That costs ballpark 2-3 million and the first order for a bunch of wafers probably also could cost 2-3 million so the initial order with nre and mask set could easily be a 5-6m cash payments (in advanc) for 28nm. Thus it's unknown at this time whether asicminer has taped out and paid the nre for the masks or for production wafers to be made. All this takes 2-3 months.

In bitcoinland many ASIC companies havnt bothered making a shuttle and have gone direct to production. Higher risk but it has usually worked. 

Tweaking/scaling up the mask happens after tape-out because without a (preliminary) physical design, sample chips can't be fabricated.  The RTL code can't go straight on the silicon!   

Tape-out is usually a cause for celebration by everyone who worked on the project, followed by trepidation awaiting the first article, the first physical samples of a chip from the manufacturing facility (semiconductor foundry). 
https://en.wikipedia.org/wiki/Tape-out

[Kipsy89]

I believe they taped out the final design of the BE300 in mid/late december and it's already submitted, the masks have been fabricated and the chips are in production right now. Come on, FC knows his job, man

[Dexter770221]

I think that tape out was done earlier. SHA256 chips are easy design (FC already designed few) and you can go straight to mass production without sampling. Sample chips were only an answer to how much these chips will be off specs and thus how much to order. FC learns on his mistakes and this time didn't ordered tons of chips that are 2 times of specs... Samples proved that chip is even better, so he could finally put a number into ordering form...

[hdbuck]

could we get it over with the "i think/hope/blah" rhetoric regarding AM?

Hellooooo FC what's up?!

Jutarul maybe? any tangible infos?

 

[aerobatic]

I believe they taped out the final design of the BE300 in mid/late december and it's already submitted, the masks have been fabricated and the chips are in production right now. Come on, FC knows his job, man

this isn't a case of friedcat not knowing what he's doing.  Of course he knows what he's doing, which is WHY he might not have placed a huge multi million dollar order at a time when bitcoin is very low priced.  He may be waiting to see what happens to the price of bitcoin, and he may also be tweaking the design, after seeing the sample chips, to make it as good as it can be before he hits the button on producing millions of them.

im not in any way accusing FC of not knowing what he's doing.  quite the contrary, he KNOWS what he's doing, so he may not have taped out yet.  The economics for spending millions making a new bitcoin mining chip are a much tougher decision today than ever in the last 18 months.

If the man in the street can buy bitcoins, at retail, for less than it costs to mine them... then you've seriously got to question whether its worth producing mining chips and other equipment.

and its a different decision for cloud mining as the economics are different if the product isn't intended to be sold at retail and can be operated in a low cost geo location (iceland etc) and host them them at scale to bring the cost down.  but then, everyone is not keen on cloud mining right now as there are too many scams and not enough evidence of actual mining going on.

[CanaryInTheMine]

I believe they taped out the final design of the BE300 in mid/late december and it's already submitted, the masks have been fabricated and the chips are in production right now. Come on, FC knows his job, man

this isn't a case of friedcat not knowing what he's doing.  Of course he knows what he's doing, which is WHY he might not have placed a huge multi million dollar order at a time when bitcoin is very low priced.  He may be waiting to see what happens to the price of bitcoin, and he may also be tweaking the design, after seeing the sample chips, to make it as good as it can be before he hits the button on producing millions of them.

im not in any way accusing FC of not knowing what he's doing.  quite the contrary, he KNOWS what he's doing, so he may not have taped out yet.  The economics for spending millions making a new bitcoin mining chip are a much tougher decision today than ever in the last 18 months.

If the man in the street can buy bitcoins, at retail, for less than it costs to mine them... then you've seriously got to question whether its worth producing mining chips and other equipment.

and its a different decision for cloud mining as the economics are different if the product isn't intended to be sold at retail and can be operated in a low cost geo location (iceland etc) and host them them at scale to bring the cost down.  but then, everyone is not keen on cloud mining right now as there are too many scams and not enough evidence of actual mining going on.

it's just a case of the usual schizophrenia...  between fc's announcements... his announcements are meds against certain forum member's schizophrenic panics and FUD.
the longer fc spends time working and not posting here (on every demand), the more numerous are the outburst of schizophrenia here... it never fails.
now, i'm not diagnosing or labeling anyone with schizophrenia, just using the word to describe crowd behavior here... so don't take it personal... just making a point...

[baristor]

sounds good ! hope i get my  moolah in time to invest in some !

[aerobatic]

I believe they taped out the final design of the BE300 in mid/late december and it's already submitted, the masks have been fabricated and the chips are in production right now. Come on, FC knows his job, man

this isn't a case of friedcat not knowing what he's doing.  Of course he knows what he's doing, which is WHY he might not have placed a huge multi million dollar order at a time when bitcoin is very low priced.  He may be waiting to see what happens to the price of bitcoin, and he may also be tweaking the design, after seeing the sample chips, to make it as good as it can be before he hits the button on producing millions of them.

im not in any way accusing FC of not knowing what he's doing.  quite the contrary, he KNOWS what he's doing, so he may not have taped out yet.  The economics for spending millions making a new bitcoin mining chip are a much tougher decision today than ever in the last 18 months.

If the man in the street can buy bitcoins, at retail, for less than it costs to mine them... then you've seriously got to question whether its worth producing mining chips and other equipment.

and its a different decision for cloud mining as the economics are different if the product isn't intended to be sold at retail and can be operated in a low cost geo location (iceland etc) and host them them at scale to bring the cost down.  but then, everyone is not keen on cloud mining right now as there are too many scams and not enough evidence of actual mining going on.

Either the price goes up eventually, or btc dies (along with associated enterprises).  Better to be sitting with loads of ready to go chips/miners when  it goes up, than to see it go up  and wait a few months for the competition to get a jump.  FC has shown foresight before, I have to assume that he still sees at least as many opportunities and scenarios as I do.
Strangely, the timing is looking as good this spring as it did last.  Let's hope for better results for AM  and BTC  this time around

its not binary.  there are plenty of other options.   for instance designing the best possible chip that could be less than half the operating cost of the present sample.  other asic designers may have demonstrated power numbers well under the 0.2 J/GH of FC's current sample, but not in the same timeframe... so FC might be thinking one of the many options is to go with what he's got as its good enough and matches some of the announced competition's spec, but equally he could be thinking its worth taking time-out while the price of btc is low and there's no hurry on the next generation to come out, and he could take that time to design a more efficient chip that might take a few more months, and that would be better suited to our present market conditions, etc.

its also incorrect to say that either the price goes up or it dies.  there's plenty of business opportunities that don't care about the price of bitcoin to succeed, and most people on these forums always say not to worry about the price of bitcoin.. for instance, one of the near-term potential businesses is the remittance market.  a specific example use case: lets say there's a nanny from the Philippines who works abroad and wants to send money home to her family.  right now its western union or some other expensive money sending options.  a specialised remittance company could utilise bitcoin transparently as the payment rails, and offer her the facility to send person to person money across country boundaries, by matching the transaction via an exchange at both ends, and allowing the nanny to send dollars in one end, and have Philippine local currency come out at the other end for their family... and yet, the nanny hasn't had to know anything about bitcoin, nor even be aware it was being used by the remittance service company... and said company has done all of that for her, and taken no currency risk nor had to hold the bitcoins for longer than milliseconds...  just long enough to do the exchange conversion at each end.  the result, much lower cost remittances without any knowledge of what bitcoin is for either sender nor receiver.

[topminingcontracts]

Did anyone have an estimate of the efficiency at the wall? Will be about 0.3W/G?

Tks