Butterfly Labs ASIC Gear Potential

[davidspitzer]

Very interesting discussion. So many possible outcomes. I will be talking to BFL in the coming weeks as I contemplate how to invest my capital. I hope they can deliver on their promises. For every difficulty the right gear at the right time will net profit. The trick is when and what.

Interested in how that conversation is going to go:

"Hello Butterfly labs, I have $XXX,XXX to spend on your awesome hardware, but I want to be the first person to receive the hardware, and I want you to hold off sending any other hardware out for 2 weeks"

"Sure thing!  But you have to give the money now as a deposit, and it's non-refundable, and we don't have solid specs yet, and we're not sure when we're shipping.   Thanks for your business!"

-EP

I am not under the impression I will be first but if there is an opportunity here it will rest with people who are the trailblazers. Conversely it is also the highest risk position as the first to the table will not have to luxury of others charting unknown waters. Risk is the very essence of a venture investment; if one is right there are big rewards if one is wrong, one pays for their mistake. This is the reason for this discussion and my subsequent discussions with BFL, so I can make an informed investment decision.

One option is to purchase the current gear with the future thought of trading it in as promised. If the new gear never materializes then the current FPGA mini rig will represent some of the best cost vs speed gear available. If the new gear does arrive most likely it will hit at the end of the year or early next year and hopefully by that time one could make back at least a good portion of their investment in the mini rigs before trade-in.

[P4man]

gyverlb,

You got me wrong, my friend. My post was not from point of view of a miner. But from point of view of an ASIC manufacturer. All I said is that they need to sell 1000 of 30,000$ rigs 1 Thps each to break even on hypothetical (educated guess) 30,000,000$ costs of full custom chip design and manufacturing. Are you going to argue my arithmetic skills here?

I would definitely argue your numbers. If by "full custom" you mean, standard cell asic (full custom stricto sensu is absolutely pointless), and assuming an affordable process, like 90 or 130nm, you are off by at least an order of magnitude. No way anyone  would plunge $30M in to a bitcoin specific chip anyway, the potential to break even on that simply isnt there.

To all the people doing the math on ROI;  redo your math assuming your minirig sc arrived as late as your fpga single, and see what happens.
 The first ones to get their hands on these might indeed see a good return, but the first ones who will get their hands on these, are...
BFL.

[davidspitzer]

gyverlb,

You got me wrong, my friend. My post was not from point of view of a miner. But from point of view of an ASIC manufacturer. All I said is that they need to sell 1000 of 30,000$ rigs 1 Thps each to break even on hypothetical (educated guess) 30,000,000$ costs of full custom chip design and manufacturing. Are you going to argue my arithmetic skills here?

I would definitely argue your numbers. If by "full custom" you mean, standard cell asic (full custom stricto sensu is absolutely pointless), and assuming an affordable process, like 90 or 130nm, you are off by at least an order of magnitude. No way anyone  would plunge $30M in to a bitcoin specific chip anyway, the potential to break even on that simply isnt there.

To all the people doing the math on ROI;  redo your math assuming your minirig sc arrived as late as your fpga single, and see what happens.
 The first ones to get their hands on these might indeed see a good return, but the first ones who will get their hands on these, are...
BFL.

Do you think that BFL will conduct mining operations themselves before releasing this gear (other then testing) Did they do (are they doing this) with their present gear?

[DILLIGAF]

They will make a lot of money back during that short window, but no where near the cost of the device. I believe, by buying this equipment, we're all just shooting each other in the foot.

Why do you think people will not make their money back. If you are one of the first to the table (timing is crucial here I think) mining at 1TH/s should recoup your $30,000 pretty quick even if the reward is halved to 25 and the difficulty starts to ramp up. Like I said calculations may be wrong but it does seem like their is an opportunity here

Considering they will release more then one at once.  Global hashrate doubles, you'll make half your guess with 1TH/s.  1 week goes by (2016 blocks found) and difficulty increases astronomically.  Goodluck!

When I plug the numbers, at Double the Difficulty and Half the reward the gear would still be repaid ($30,000) in 30 days

Did you account for the difficulty rising twice (possibly 3 or even 4 times) during that period?

When companies order chips like these, they arrive all at once..  If BFL has any common sense, they will flood the market with these things so everyone has a 'chance' at getting the first 1TH/s machine.  Then everyone will complain that their ROI has risen back to 2 years.

1 month gross rates for 1TH/s at $6.38 exchange rate

At present difficulty and 25 reward  - $62850.86 USD
Double the Difficulty 25 reward - $31425.43 USD
Triple the Difficulty 25 reward - $20950.29 USD
Quadruple the Difficulty 25 reward - $15712.71 USD

It still seems to me that even as the difficulty ramps up if you are one of the first to market that you will enjoy a good rate of return until the field equalizes. So your actual return over the first 6 months would probably be a sliding scale of upward difficulty but your initial investment seems like it should be paid off rather quickly

I think it would be safe to think that they can get at least two boxes a day out the door in the first week so when they come online.

First week present diff $6 BTC.

฿635.31/day for $3,811.88/day $26,683.13/week

Now in seven days the diff will have doubled so second week you can halve those figures in this you can make your money back, more likely to be the case at least four plus boxes a day shipping under that idea the first 3.5 days present diff $6 BTC.

฿635.31/day for $3,811.88/day $13,341.58/for 3.5 days

Now the diff will have gone up by four times so every day after this you get one quarter (฿158.82/day) the return. Not so likely to get your money back now if they manage to get eight boxes a day out the door the same 3.5 days above applies as well but then in another 3.5 days the diff goes up four times again so your return drops by another quarter a day (฿39.7/day) at the end of the first week. To end up at the point where there is not a hope in hell of getting your money back. Nothing in these figures takes into account the doubling or quadrupling of the coins coming onto to the market for this brief time while the system readjusts to the new hashing capacity nor the inevitable extra hashing that will continue to be added as the days/weeks/months go by, as they are not going to want to sell only 50-60 of these 1th/s boxes let alone none of the other offerings they have listed. As well they have end of October listed for availability if they slip a month then we are into the reward halving territory then the best case payback I list here turns into not a hope in hell of one.

[P4man]

Do you think that BFL will conduct mining operations themselves before releasing this gear (other then testing) Did they do (are they doing this) with their present gear?

They burn in their minirigs, which to be fair, is what youd expect anyway. You know, testing. But its not done on testnet.
Now their current hardware doesnt have a significant impact on the network, but these asics will change that. Whether or not they will mine for profit; who knows? But they sure wont tell you if they do. Nor will they tell you who gets those rigs first.

[cmg5461]

I think it would be safe to think that they can get at least two boxes a day out the door in the first week so when they come online.

First week present diff $6 BTC.

฿635.31/day for $3,811.88/day $26,683.13/week

Now in seven days the diff will have doubled so second week you can halve those figures in this you can make your money back, more likely to be the case at least four plus boxes a day shipping under that idea the first 3.5 days present diff $6 BTC.

฿635.31/day for $3,811.88/day $13,341.58/for 3.5 days

Now the diff will have gone up by four times so every day after this you get one quarter (฿158.82/day) the return. Not so likely to get your money back now if they manage to get eight boxes a day out the door the same 3.5 days above applies as well but then in another 3.5 days the diff goes up four times again so your return drops by another quarter a day (฿39.7/day) at the end of the first week. To end up at the point where there is not a hope in hell of getting your money back. Nothing in these figures takes into account the doubling or quadrupling of the coins coming onto to the market for this brief time while the system readjusts to the new hashing capacity nor the inevitable extra hashing that will continue to be added as the days/weeks/months go by, as they are not going to want to sell only 50-60 of these 1th/s boxes let alone none of the other offerings they have listed. As well they have end of October listed for availability if they slip a month then we are into the reward halving territory then the best case payback I list here turns into not a hope in hell of one.

https://docs.google.com/spreadsheet/ccc?key=0Ai2b4IJuKUZhdE5EVFJkUF95TS1BVjdxaEFsSVlQSkE

check out my calculator

[EPiSKiNG]

Very interesting discussion. So many possible outcomes. I will be talking to BFL in the coming weeks as I contemplate how to invest my capital. I hope they can deliver on their promises. For every difficulty the right gear at the right time will net profit. The trick is when and what.

Interested in how that conversation is going to go:

"Hello Butterfly labs, I have $XXX,XXX to spend on your awesome hardware, but I want to be the first person to receive the hardware, and I want you to hold off sending any other hardware out for 2 weeks"

"Sure thing!  But you have to give the money now as a deposit, and it's non-refundable, and we don't have solid specs yet, and we're not sure when we're shipping.   Thanks for your business!"

-EP

I am not under the impression I will be first but if there is an opportunity here it will rest with people who are the trailblazers. Conversely it is also the highest risk position as the first to the table will not have to luxury of others charting unknown waters. Risk is the very essence of a venture investment; if one is right there are big rewards if one is wrong, one pays for their mistake. This is the reason for this discussion and my subsequent discussions with BFL, so I can make an informed investment decision.

One option is to purchase the current gear with the future thought of trading it in as promised. If the new gear never materializes then the current FPGA mini rig will represent some of the best cost vs speed gear available. If the new gear does arrive most likely it will hit at the end of the year or early next year and hopefully by that time one could make back at least a good portion of their investment in the mini rigs before trade-in.

Quite eloquently put.  Good luck sir!  I wish you the best of all possible outcomes.

-EP

[mc_lovin]

nice spreadsheet they should sell these at auction, one at a time, and take the profits and just hold onto them as a reserve so their company gets more money in their pocket and people have a chance at getting each unit, and instead of giving someone a unit that will make them their investment back in a few days and then raping the network forever, this way they bid for it and might end up paying $150,000 instead of $40,000 to get it.  

[davidspitzer]

Very interesting discussion. So many possible outcomes. I will be talking to BFL in the coming weeks as I contemplate how to invest my capital. I hope they can deliver on their promises. For every difficulty the right gear at the right time will net profit. The trick is when and what.

Interested in how that conversation is going to go:

"Hello Butterfly labs, I have $XXX,XXX to spend on your awesome hardware, but I want to be the first person to receive the hardware, and I want you to hold off sending any other hardware out for 2 weeks"

"Sure thing!  But you have to give the money now as a deposit, and it's non-refundable, and we don't have solid specs yet, and we're not sure when we're shipping.   Thanks for your business!"

-EP

I am not under the impression I will be first but if there is an opportunity here it will rest with people who are the trailblazers. Conversely it is also the highest risk position as the first to the table will not have to luxury of others charting unknown waters. Risk is the very essence of a venture investment; if one is right there are big rewards if one is wrong, one pays for their mistake. This is the reason for this discussion and my subsequent discussions with BFL, so I can make an informed investment decision.

One option is to purchase the current gear with the future thought of trading it in as promised. If the new gear never materializes then the current FPGA mini rig will represent some of the best cost vs speed gear available. If the new gear does arrive most likely it will hit at the end of the year or early next year and hopefully by that time one could make back at least a good portion of their investment in the mini rigs before trade-in.

Quite eloquently put.  Good luck sir!  I wish you the best of all possible outcomes.

-EP

In any event my wife promised not to kill me if I end up being wrong

[cmg5461]

nice spreadsheet they should sell these at auction, one at a time, and take the profits and just hold onto them as a reserve so their company gets more money in their pocket and people have a chance at getting each unit, and instead of giving someone a unit that will make them their investment back in a few days and then raping the network forever, this way they bid for it and might end up paying $150,000 instead of $40,000 to get it.  

awesome idea.  people will pay the premium to have it early.  Like releasing 2-3 a week or something.  eventually the price will trickle down to that break even point, and BFL will make more money than they planned on.  Why sell 5 for 30k when you can sell 1 for 150k while it's in high demand?

[bulanula]

Do you think that BFL will conduct mining operations themselves before releasing this gear (other then testing) Did they do (are they doing this) with their present gear?

They burn in their minirigs, which to be fair, is what youd expect anyway. You know, testing. But its not done on testnet.
Now their current hardware doesnt have a significant impact on the network, but these asics will change that. Whether or not they will mine for profit; who knows? But they sure wont tell you if they do. Nor will they tell you who gets those rigs first.

It is simple mate : the one who puts the most money at the earliest moment gets to be the first.

Think gigavps or Inaba or other fanboys.

Who hands over the most money right now wins the game !

[MrTeal]

Now, what would you consider your break even point if you have to give BFL money at the beginning of July, and you receive your SC Minirig in December? Would you still consider that 30 days to recoup your investment?

How long between the first payments for Singles and the first Singles in the wild?
How long between the first preorders for Rigboxes and the first Minirigs in the wild?

[malevolent]

IMO they are just bluffing

I can't believe people are falling for this... remember the ''1.05GH/s and 19.6W for $500''?

Either this is a long prepared scam or again incorrect simulation results (or they bribed their way through government-owned wafer factory in some developing country).

An ASIC performing at the speed of 1300 HD5970s (about $600k) for $30k? Who is going to fall for that? With a price of, say, $150k for 1TH/s ASIC they would appear at least more credible and still remain competitive. This doesn't make sense. If by chance they would release it with currently advertised specs and price this would mean end to GPU mining as has happened with CPU mining over a year ago.

[mc_lovin]

The reason I said earlier that this would doom bitcoin is because right now, the "mining process" is done by the individual users in a "decentralized" fashion, and your average Joe can get a video card and start earning bitcoins, and then decide wether to sell them or stash them, if along comes an ASIC that is thousands of times more powerful, it makes GPU mining obsolete and all the average Joes out there are going to flip their cards on eBay (and the market will flood with GPUs, lowering their value) and people will get out of the mining game, and in a lot of cases, get out of bitcoin altogether.  The thousands of users right now who are mining bitcoin on their computers are likely to purchase things off each other with those coins, and if you took away their income overnight, then a large percentage of them will walk away feeling cheated as some monster miner is mining all the coins.

Satoshi would not want it to come to this.

Forgive me if this logic doesn't make sense.  Personally, if I could no longer mine coins anymore, I would sell all my hardware (as it is now useless) and then probably not invest in bitcoin anymore, unless those 3.5GH/s for $150 units are in large enough supply for me to get some, but in that case everyone will have them...

[gyverlb]

I would definitely argue your numbers. If by "full custom" you mean, standard cell asic (full custom stricto sensu is absolutely pointless), and assuming an affordable process, like 90 or 130nm, you are off by at least an order of magnitude. No way anyone  would plunge $30M in to a bitcoin specific chip anyway, the potential to break even on that simply isnt there.

$3M for the design may even be higher than needed. The ASIC wouldn't be nearly as complex as a CPU or a DSP for example, so comparing the R&D cost to CPU/DSP design isn't really meaningful.

Is someone aware of R&D cost examples for simple ASIC designs ? Crypto-accelerators could be in the same ballpark although a bit more complex I guess.

[rjk]

government-owned wafer factory in some developing country

LOL

Satoshi would not want it to come to this.

How do you know? Are you channeling him now?

[2112]

If by "full custom" you mean, standard cell asic (full custom stricto sensu is absolutely pointless),

Why you think that full custom design is absolutely pointless?

We have already experienced that the Bitcoin hashing requires far out of ordinary chip design:

1) unusually high toggle rate
2) problems with supplying enough power
3) problems with dissipating enough heat
4) simultaneous switching noise problem with internal signal immunity violations

On the other hand Bitcoin hashing is highly regular. Once those problems are solved for single round most of the full design will consist of the copies of said single round.
With this respect the full custom design should be worthwhile. Moreover, once well designed Bitcoin hasher could become a standard benchmark case for comparison of ASIC processes. It also has a nice nearly-self-testing property: there are no unreachable internal states and almost 100% of faults show in the output for no additional effort.

Maybe you were thinking about some financials?  The overall design is highly repetitive which greatly reduces the total cost.

[mc_lovin]

Satoshi would not want it to come to this.

How do you know? Are you channeling him now?

I just think he would have wanted "the people" to have the coins, not the millionaires who can afford these special mining machines.  So what if the Federal Reserve bought a bunch?  If they are only $40k for a 1 TH/s unit, what is stopping them from dropping $10 million on 250 of these units and having 250 TH/s?  And what's stopping them from doing that over and over again until they control the supply of bitcoins?  Now, that would increase the value of the bitcoins that are already out in the wild, but we have to be aware of the fact that it allows the upper 1% TOO much control.

[rjk]

Satoshi would not want it to come to this.

How do you know? Are you channeling him now?

I just think he would have wanted "the people" to have the coins, not the millionaires who can afford these special mining machines.  So what if the Federal Reserve bought a bunch?  If they are only $40k for a 1 TH/s unit, what is stopping them from dropping $10 million on 250 of these units and having 250 TH/s?  And what's stopping them from doing that over and over again until they control the supply of bitcoins?  Now, that would increase the value of the bitcoins that are already out in the wild, but we have to be aware of the fact that it allows the upper 1% TOO much control.

I'm not a millionaire, and I can afford a Jalapeno device for $150. I'm sure lots of others can too.

[mc_lovin]

Satoshi would not want it to come to this.

How do you know? Are you channeling him now?

I just think he would have wanted "the people" to have the coins, not the millionaires who can afford these special mining machines.  So what if the Federal Reserve bought a bunch?  If they are only $40k for a 1 TH/s unit, what is stopping them from dropping $10 million on 250 of these units and having 250 TH/s?  And what's stopping them from doing that over and over again until they control the supply of bitcoins?  Now, that would increase the value of the bitcoins that are already out in the wild, but we have to be aware of the fact that it allows the upper 1% TOO much control.

I'm not a millionaire, and I can afford a Jalapeno device for $150. I'm sure lots of others can too.

Yeah good point, I guess it depends on the volume they are able to make those...  If we can get them in stacks, I guess the whole world is going to mine with Jalapeno's this winter..  So much for heating my house this winter with my rigs!  The difficulty/trading price/network hashrate is going to go bonkers for awhile but I suppose it will correct itself...