Butterfly Labs invests heavily in high speed production equipment

[Phinnaeus Gage]

eBay seller mynhung2006 purchased a lot of things last year that BFL could possibly use in their labs. Since he's having a fire sale, I would suggest re-calling the seller to see if the following item he purchased is next up on the auction block.



http://feedback.ebay.com/ws/eBayISAPI.dll?ViewFeedback2&userid=mynhung2006&ftab=FeedbackLeftForOthers

Great seller, shrink warpped unit and help load it for free, will buy from again   Seller: moonray ( 2927)     May-17-11 18:29
    -- (#260780101626)

I ain't got the foggiest idea why the eBay seller stated that he's selling these items for some lady, when in fact he purchased what he's selling.

~Bruno~

[neotrino]

My concern level went from yellow to orange when they raised  power their chips by 50%.
My concern level went from orange to red when I heard about the lifetime warranty on a product that was never tested. Any serious company says it's suicide.

BFL's PR has become so colored as to be impossible.
No one could ever raise the power chips of 50% - month before the sale, without increasing power consumption and without extreme overclocking. Rumors about firmware is funny, multicore is a very advanced and expensive technology. BFL offers what can not offer big companies such as Intel, NVidia, AMD.... This makes the BFL is for me an incredible company. Of course, I wish them luck but I'm not crazy enough to buy preorder. I'll wait until they have a finished product. I am also concerned about creating PR using fake photos. I am also shocked unfriendly treatment of customers who ask too many questions - they still want to invest a lot of money, sometimes their life savings, they just want to be careful. I am also worried that a month before shipment, they are not have  finished product. Where is the time for testing? How can you sell untested device? 24 hours is not enough.  Every serious company will be  laughing of this . I was also worried that the average time to develop ASIC chips (all  phases) takes an average of 18 months, as a small company you are working suspiciously fast. People have a lot of concerns, you as a company should try to resolve this instead of unkindly reply.

Perhaps you are missing the point that their previous specs were only guessed (in a very conservative way). And until they got the first ASIC assembled they didn't managed to know about the real specs they would get.

Assembling an ASIC is very expensive, so you can't assemble a "beta" version of the chip just for testing. You have to assemble the final production-ready chip and therefore you don't know the real specs until the final stages of production

Specification of chips is well known in the design phase. Well known is  number of logic gates and technological process - it is easy to calculate the maximum operating frequency at  maximum safe temperature. Calculation error may amount to a few % , not 50% !!!
This is not FPGA  so you can just buy ASIC in a store and then check  performance. ASICs are built from scratch. If I'm wrong correct me.

I guess such calculations are not so easy to do, and depend a lot on the manufacturing process (Ex: 40nm vs 28nm). So probably they didn't have done right until now. It could be even possible that the original calculations were done with a very conservative manufacturing process (40nm) and finally they managed, to manufacture the chips with a better technology (28nm).

[squid]

I guess such calculations are not so easy to do, and depend a lot on the manufacturing process (Ex: 40nm vs 28nm). So probably they didn't have done right until now. It could be even possible that the original calculations were done with a very conservative manufacturing process (40nm) and finally they managed, to manufacture the chips with a better technology (28nm).

You are dreaming if you think BFL (or anyone else for that matter) is using 28 or 40 nm technology.

[bitmar]

My concern level went from yellow to orange when they raised  power their chips by 50%.
My concern level went from orange to red when I heard about the lifetime warranty on a product that was never tested. Any serious company says it's suicide.

BFL's PR has become so colored as to be impossible.
No one could ever raise the power chips of 50% - month before the sale, without increasing power consumption and without extreme overclocking. Rumors about firmware is funny, multicore is a very advanced and expensive technology. BFL offers what can not offer big companies such as Intel, NVidia, AMD.... This makes the BFL is for me an incredible company. Of course, I wish them luck but I'm not crazy enough to buy preorder. I'll wait until they have a finished product. I am also concerned about creating PR using fake photos. I am also shocked unfriendly treatment of customers who ask too many questions - they still want to invest a lot of money, sometimes their life savings, they just want to be careful. I am also worried that a month before shipment, they are not have  finished product. Where is the time for testing? How can you sell untested device? 24 hours is not enough.  Every serious company will be  laughing of this . I was also worried that the average time to develop ASIC chips (all  phases) takes an average of 18 months, as a small company you are working suspiciously fast. People have a lot of concerns, you as a company should try to resolve this instead of unkindly reply.

Perhaps you are missing the point that their previous specs were only guessed (in a very conservative way). And until they got the first ASIC assembled they didn't managed to know about the real specs they would get.

Assembling an ASIC is very expensive, so you can't assemble a "beta" version of the chip just for testing. You have to assemble the final production-ready chip and therefore you don't know the real specs until the final stages of production

Specification of chips is well known in the design phase. Well known is  number of logic gates and technological process - it is easy to calculate the maximum operating frequency at  maximum safe temperature. Calculation error may amount to a few % , not 50% !!!
This is not FPGA  so you can just buy ASIC in a store and then check  performance. ASICs are built from scratch. If I'm wrong correct me.

I guess such calculations are not so easy to do, and depend a lot on the manufacturing process (Ex: 40nm vs 28nm). So probably they didn't have done right until now. It could be even possible that the original calculations were done with a very conservative manufacturing process (40nm) and finally they managed, to manufacture the chips with a better technology (28nm).

so, if they have not yet ready chips, how these (5Gh/chip) calculations come from? 40-28nm  is impossible for such a small company (and very expensive). Is it possible that mistake in the calculation of the 50% not even having chips? I do not understand. Larger companies are able to evaluate exactly the power of designed chips.

[neotrino]

I guess such calculations are not so easy to do, and depend a lot on the manufacturing process (Ex: 40nm vs 28nm). So probably they didn't have done right until now. It could be even possible that the original calculations were done with a very conservative manufacturing process (40nm) and finally they managed, to manufacture the chips with a better technology (28nm).

You are dreaming if you think BFL (or anyone else for that matter) is using 28 or 40 nm technology.

Why you say that is impossible? Nowadays practically all semiconductor manufacturing plants have technology to manufacturer chips with 28nm.

Of course they are not building the chips (that's just impossible) they are outsourcing the manufacturing process to companies like TSMC

For example Altera (a very well-know FPGA manufacturer) offer a service that allows you to copy your design from an Altera FPGA to an ASIC with 28nm or 40nm https://en.wikipedia.org/wiki/Altera#ASICs

http://www.altera.com/devices/asic/hardcopy-asics/about/hrd-index.html

[squid]

I guess such calculations are not so easy to do, and depend a lot on the manufacturing process (Ex: 40nm vs 28nm). So probably they didn't have done right until now. It could be even possible that the original calculations were done with a very conservative manufacturing process (40nm) and finally they managed, to manufacture the chips with a better technology (28nm).

You are dreaming if you think BFL (or anyone else for that matter) is using 28 or 40 nm technology.

Why you say that is impossible? Nowadays practically all semiconductor manufacturing plants have technology to manufacturer chips with 28nm.

Of course they are not building the chips (that's just impossible) they are outsourcing the manufacturing process to companies like TSMC

For example Altera (a very well-know FPGA manufacturer) offer a service that allows you to copy your design from an Altera FPGA to an ASIC with 28nm or 40nm https://en.wikipedia.org/wiki/Altera#ASICs

http://www.altera.com/devices/asic/hardcopy-asics/about/hrd-index.html

If NRE is on the order of >$100,000 USD for 130 nm ASICs, how can you expect BFL to afford 28/40nm type stuff that only the largest of companies can afford?

[neotrino]

I guess such calculations are not so easy to do, and depend a lot on the manufacturing process (Ex: 40nm vs 28nm). So probably they didn't have done right until now. It could be even possible that the original calculations were done with a very conservative manufacturing process (40nm) and finally they managed, to manufacture the chips with a better technology (28nm).

You are dreaming if you think BFL (or anyone else for that matter) is using 28 or 40 nm technology.

Why you say that is impossible? Nowadays practically all semiconductor manufacturing plants have technology to manufacturer chips with 28nm.

Of course they are not building the chips (that's just impossible) they are outsourcing the manufacturing process to companies like TSMC

For example Altera (a very well-know FPGA manufacturer) offer a service that allows you to copy your design from an Altera FPGA to an ASIC with 28nm or 40nm https://en.wikipedia.org/wiki/Altera#ASICs

http://www.altera.com/devices/asic/hardcopy-asics/about/hrd-index.html

If NRE is on the order of >$100,000 USD for 130 nm ASICs, how can you expect BFL to afford 28/40nm type stuff that only the largest of companies can afford?

I don't know how much money BFL was able to raise, but 1M sounds feasible to me.

Of course I am only speculating, I don't know about real numbers.

Probably some of you will find this page useful http://opencores.org/forum,OpenRISC%20-%20ASIC%20Funding,0,4358 where OpenRISC talks about minimum orders for different ASIC manufacturing process.


The point of the discussion was about how they managed to bump the original specifications up to a 50% more while keeping the original cost.

And, perhaps is because of this, they originally targeted one manufacturing process of XXXnm, and they finally raised more money than expected so they were able to go for YYYnm process which allowed them to boost the performance and lower the power consumption.

It would be nice to know how much nm will use the manufacturing process of the BF ASICs.

Regards

[Bogart]

It would be nice to know how much nm will use the manufacturing process of the BF ASICs.

I want the one with the smaller Enn Emms¹.

[neotrino]

It would be nice to know how much nm will use the manufacturing process of the BF ASICs.

I want the one with the smaller Enn Emms¹.

Hoping they will answer:

https://forums.butterflylabs.com/showthread.php/95-About-ASIC-manufacturing-process-XXXnm

[PuertoLibre]

It would be nice to know how much nm will use the manufacturing process of the BF ASICs.

I want the one with the smaller Enn Emms¹.

Hoping they will answer:

https://forums.butterflylabs.com/showthread.php/95-About-ASIC-manufacturing-process-XXXnm

You should also ask how many hashing units they have per chip. As well as....if any are disabled (or were disabled) due to firmware.

[RHA]

It would be nice to know how much nm will use the manufacturing process of the BF ASICs.

Hoping they will answer:
https://forums.butterflylabs.com/showthread.php/95-About-ASIC-manufacturing-process-XXXnm

I bet they won't reveal this yet. I wouldn't reveal. Near all buyers don't care and the competitors could have an edge.

[franky1]

instead of highlighting the errors and be rewarded with inaba replies of tin foil hat troll, instead of actual answers. i will not mention the errors already mentioned, but instead just ask some simple questions.

1. does BFL have a finished ASIC prototype that they are testing right now ready to mass/batch produce once quality control checks are done

2. has BFL already done part of question one and already batch producing finalised ASICS.

3. if yes to question 2 are any of them currently going through a burn in process either solo mining or pool mining. and if so which?

4. does BFL have any images of CURRENT workshop equipment that are not third party sources (ebay/google images)

[bitmar]

It would be nice to know how much nm will use the manufacturing process of the BF ASICs.

Hoping they will answer:
https://forums.butterflylabs.com/showthread.php/95-About-ASIC-manufacturing-process-XXXnm

I bet they won't reveal this yet. I wouldn't reveal. Near all buyers don't care and the competitors could have an edge.

you're right, we will not know.

https://forums.butterflylabs.com/showthread.php/95-About-ASIC-manufacturing-process-XXXnm
"BFL_Josh:
We will be releasing that information within the next few weeks. "

My suspicion growing. Business secret.... This is an easy way to fool people. But is also reasonable. whatever they should authenticate their actions for their own good. Or they gathered enough money to not care.... or they believe people are so naive...and as history shows they are right.

[yrtrnc]

What I dont get is that they are using a machine that costs 4,499 usd on eBay to manufacture millions of dollars wort of goods??!!
http://www.ebay.com/itm/ESSEMTEC-RO300FC-REFLOW-OVEN-/320989507198?pt=LH_DefaultDomain_0&hash=item4abc772e7e&nma=true&si=OJ7anbJ%2FAqbSVi60eILcRUvNe54%3D&orig_cvip=true&rt=nc&_trksid=p2047675.l2557

Can someone pls explain this?

[Bogart]

To contrast, ASICMINER has already stated that they will be using a 130nm process, and ngzhang/Avalon stated from the outset that 130nm or 110nm will be used.  I don't think cablepair/bASIC has disclosed their process size.

[Bogart]

What I dont get is that they are using a machine that costs 4,499 usd on eBay to manufacture millions of dollars wort of goods??!!
http://www.ebay.com/itm/ESSEMTEC-RO300FC-REFLOW-OVEN-/320989507198?pt=LH_DefaultDomain_0&hash=item4abc772e7e&nma=true&si=OJ7anbJ%2FAqbSVi60eILcRUvNe54%3D&orig_cvip=true&rt=nc&_trksid=p2047675.l2557

Can someone pls explain this?

What's not to understand?  This machine just performs one part of the process.  The parts that go into the machine still have to come from somewhere, and there is a lot more processing involved besides just running the units through the reflow oven.

It's not like the replicators on Star Trek where they just say "make me 9,000 BitForce SC products" and it fabricates them out of thin air.

[fcmatt]

What I dont get is that they are using a machine that costs 4,499 usd on eBay to manufacture millions of dollars wort of goods??!!
http://www.ebay.com/itm/ESSEMTEC-RO300FC-REFLOW-OVEN-/320989507198?pt=LH_DefaultDomain_0&hash=item4abc772e7e&nma=true&si=OJ7anbJ%2FAqbSVi60eILcRUvNe54%3D&orig_cvip=true&rt=nc&_trksid=p2047675.l2557

Can someone pls explain this?

A small business has to manage their pennies properly or fail. Buying on ebay to save money is reasonable and smart if it wirks out.

[Zeek_W]

instead of highlighting the errors and be rewarded with inaba replies of tin foil hat troll, instead of actual answers. i will not mention the errors already mentioned, but instead just ask some simple questions.

1. does BFL have a finished ASIC prototype that they are testing right now ready to mass/batch produce once quality control checks are done

2. has BFL already done part of question one and already batch producing finalised ASICS.

3. if yes to question 2 are any of them currently going through a burn in process either solo mining or pool mining. and if so which?

4. does BFL have any images of CURRENT workshop equipment that are not third party sources (ebay/google images)

Very valid questions!

[squeept]

What I dont get is that they are using a machine that costs 4,499 usd on eBay to manufacture millions of dollars wort of goods??!!
http://www.ebay.com/itm/ESSEMTEC-RO300FC-REFLOW-OVEN-/320989507198?pt=LH_DefaultDomain_0&hash=item4abc772e7e&nma=true&si=OJ7anbJ%2FAqbSVi60eILcRUvNe54%3D&orig_cvip=true&rt=nc&_trksid=p2047675.l2557

Can someone pls explain this?

What does this have to do with ANYTHING?

I've used a $100 multimeter that my dad gave to me to repair hundreds of thousands of dollars worth of electronics. Therefore, I am doing something wrong?

[Phinnaeus Gage]

What I dont get is that they are using a machine that costs 4,499 usd on eBay to manufacture millions of dollars wort of goods??!!
http://www.ebay.com/itm/ESSEMTEC-RO300FC-REFLOW-OVEN-/320989507198?pt=LH_DefaultDomain_0&hash=item4abc772e7e&nma=true&si=OJ7anbJ%2FAqbSVi60eILcRUvNe54%3D&orig_cvip=true&rt=nc&_trksid=p2047675.l2557

Can someone pls explain this?

What's not to understand?  This machine just performs one part of the process.  The parts that go into the machine still have to come from somewhere, and there is a lot more processing involved besides just running the units through the reflow oven.

It's not like the replicators on Star Trek where they just say "make me 9,000 BitForce SC products" and it fabricates them out of thin air.

And don't forget, each of those 9,000 units will endure a minimum 12 hour endurance test prior to shipping. Somebody else can do the math on how many bitcoins can be mined in 4,500+ hours with that many BitForce SC.

@Energizer: We do test all the units for a minimum of 12 hours prior to shipping. This is our endurance test
that is done in a harsh environment to make sure the unit is stable and will not encounter any problem in its
lifetime.

Regards,

~Bruno~