[VIDEO] Butterfly Labs (BFL) Bitforce SC ASIC Test

[jedunnigan]

Saw this on unofficial IRC. I'll let you guys pick it apart.

https://www.youtube.com/watch?v=4C4bgho5JSI

Can you embed on here? Is it that flash tag?!

[Bowjob]

Great. They are now in the prototype stage.

[Frankie Delaney]

No hashing. Just testing, useless video.

[Miner99er]

Yeah, a couple million hashes on testnet or something would be more than fair, shouldn't take that long to do it. This video didn't show ANYTHING of substance.

[Bogart]

42W at idle?  Good lawd.  My x6500 FPGAs only burns up 17.2W at full load...

Didn't they like...put a dummy load in place of the ASIC chips and actually test these boards or something while they were waiting on chip delivery?

I guess not.

[johnyj]

What component can draw 42W at idle?

I think they should provide an ATX PSU power supply converter where many standard 12V connectors can be used

[dogie]

HAHAHAHAHAHAHAHAHAHAHAHAHA *dies*

So they've tested that the board can handle the data. They are still to:

Actually have a chip
Actually make the chip work
Actually bin the chip
Sort out the communication protocols, never mind proper USB interfaces etc
Get the power consumption down to ANYTHING near what they promised
Actually make it hash
Actually make it hash to the required speed
Actually make it hash consistently
Actually make it hashable 24/7
Actually prevent it overheating [That's 160W TDP on the equivalent of a dual slot GPU, which would typically warrant ~40CFM - they have 0, AND will be placing many in a an enclosed case]
Work out how to do the power distribution
Make the shit fit in the cases.
Work out sufficient testing algorithms and the logistics of burning in '300 units a day'
Sort out the logistics of labelling, warranties, who gets what.
Actually deliver anything that works for more than a few days.

ETA 5 months, GG.

[GenTarkin]

Im confused....
Now, Im not that versed in power circuitry ... but how the hell can the PCB power circuitry design be so shitty that the board itself is using nearly 2x the power than the chip? iirc the SC was supposed to be like 60watts right? The whole unit is using 150ish .. how the fuck can the board be using 90watts?!?!?
Did some noob engineer design the board or some shit? I mean really.

[BitSyncom]

Yay! they are not a scam at least. This means they'll eventually ship a product.

fyi: we reached that stage 1.2 month before we shipped. Let's see how they handle their remainder issues.

[S M I L Y]

Im surprised that their VR solution can even handle the wattage!!!   They were targeting 60w!    They must have known they where going to exceed their power budget!

[johnyj]

Yay! they are not a scam at least. This means they'll eventually ship a product.

fyi: we reached that stage 1.2 month before we shipped. Let's see how they handle their remainder issues.

this

[DobZombie]

Saw this on unofficial IRC. I'll let you guys pick it apart.

https://www.youtube.com/watch?v=4C4bgho5JSI

Can you embed on here? Is it that flash tag?!

I feel I speak for a lot of the community here when I say...

HOO-BLOODY-RAY!!!

[iCEBREAKER]

Is it just me, or do the lights dim when he flips on the BetterFail thingy?

[sunblaster]

does it even MH/s!

[Inspector 2211]

does it even MH/s!

The more important question is:
WILL IT BLEND?

[tvbcof]

Yay! they are not a scam at least. This means they'll eventually ship a product.

fyi: we reached that stage 1.2 month before we shipped. Let's see how they handle their remainder issues.

Really?  If they were a scam you could expect as realistic a 'prototype test' as they could muster, and I saw nothing in that vid which proved much of anything.  It amazes me to see how many people consider a vid of some numbers changing on a screen to be 'proof' of anything more than a tiny amount of scripting ability.

You could also expect that they would make the 'test' look just good enough to give people hope (and thus not cancel their pre-orders) but bad enough to have a reason for another bunch of delays.

The only thing which gives me some amount of hope is that they made things look worse than they really needed to.

As I said for Avalon, I won't believe with a high degree of certainty that they can hash unless they can publish a set of hashes within {x} amount of time of a known dataset being globally available.  And even then, not until a very credible third-party is standing over their shoulder to make sure that they don't have an Avalon hiding under the desk or something of that nature.  I wouldn't trust these people at all.

[Rawted]

Looks like some charity is about to make out with $100k.

https://forums.butterflylabs.com/content/123-bfl-offers-1000-btc-charity-if-they-miss-their-power-targets.html

[Tea, Towels, Vogons etc.]

Well, they had some type of chip(s) in that board, because it drew almost 160w when they ran the test.

[dogie]

Well, they had some type of chip(s) in that board, because it drew almost 160w when they ran the test.

Others were right in suggesting it was likely a dummy load for them to test the power distribution. Could literally just put a series of resistors under the heatsink and manually ram current through.

[smoothie]

Yay! they are not a scam at least. This means they'll eventually ship a product.

fyi: we reached that stage 1.2 month before we shipped. Let's see how they handle their remainder issues.

Good one Yifu!  I LOLed.

[techwtf]

Anyone can tell me:
1 why only 15 nonce like response is returned?
2 what was he tested? single chip? hash core?
3 why it take ~20s to have the power to return idle?

[BBQKorv]

Oh my god.

So much fail in one place.

Where is the no significant shipping in May bet?

That would be a really nice bet to make some money. The right horse should be easy to pick 

[Meizirkki]

The only party surprised by higher power usage appears to be BFL itself.. even after they failed to do correct estimates for their FPGAs.

<insert history repetition quote>

[mobodick]

Saw this on unofficial IRC. I'll let you guys pick it apart.

https://www.youtube.com/watch?v=4C4bgho5JSI

Can you embed on here? Is it that flash tag?!

I feel I speak for a lot of the community here when I say...

HOO-BLOODY-RAY!!!

You're hashing prematurely.

[mobodick]

Anyone can tell me:
1 why only 15 nonce like response is returned?
2 what was he tested? single chip? hash core?
3 why it take ~20s to have the power to return idle?

It looks much more as if they are just testing the controller IC of the board and not the actual ASIC chip.

[willphase]

Anyone can tell me:
1 why only 15 nonce like response is returned?
2 what was he tested? single chip? hash core?
3 why it take ~20s to have the power to return idle?

1. sounds one one of the cores isn't working.
2. single chip on a board - each chip has 16 cores (see image)
3. presumably the test sends a block header (or equivalent) to each core and gets them to iterate over the nonce until the hash < difficulty - they program difficulty in - and the fact that they all came up with the same nonce means they are hashing.  It's the smallest unit of work that an ASIC that does SHA1-256(SHA1-256(data)) would have to do to prove it works.

Basically the video proves that chip works on a board, but only 15 of the 16 cores are hashing, and it uses a lot more power than expected (approx 4x)

Will

[mobodick]

Anyone can tell me:
1 why only 15 nonce like response is returned?
2 what was he tested? single chip? hash core?
3 why it take ~20s to have the power to return idle?

1. sounds one one of the cores isn't working.
2. single chip on a board - each chip has 16 cores (see image)
3. presumably the test sends a block header (or equivalent) to each core and gets them to iterate over the nonce until the hash < difficulty - they program difficulty in - and the fact that they all came up with the same nonce means they are hashing.  It's the smallest unit of work that an ASIC that does SHA1-256(SHA1-256(data)) would have to do to prove it works.

Basically the video proves that chip works on a board, but only 15 of the 16 cores are hashing, and it uses a lot more power than expected (approx 4x)

Will

Aah, thanks for the clarification.

[Dexter770221]

Did I heard "board with extra regulators"? So, problem is in chips not the board. Theydraw to much power. Same story as year ago with SC single. Promised 20W, delivered 80W. Now 60-160, it's progress. With 4th generation maybe they learn how to properly estimate power consumption.

[Frizz23]

Yay! they are not a scam at least. This means they'll eventually ship a product.

fyi: we reached that stage 1.2 month before we shipped. Let's see how they handle their remainder issues.

Based on their track record of failures it will take them at least 5 times of what it took you guys. So we can expect a non-prototype product that works (more or less) within specs in about 1/2 year.

[smoothie]

Josh: We are testing our "chips" and they hash at X rate but should hash at Y rate...could be the boards, the chips, or the clockbuffers...we don't know.

Investor: So when are you guys shipping? Next week? Two weeks?

Josh: We hope/should be shipping in two weeks.

Investor: Will I make profit with my order via mining?

Josh: You "should" profit.

Investor: Yay! I might profit after giving you a ton of money I could have made at least 1000% on in the same time it took you to develop your "asic" device. I love BFL. They are my hero.

 

[Bogart]

Did I heard "board with extra regulators"? So, problem is in chips not the board. Theydraw to much power. Same story as year ago with SC single. Promised 20W, delivered 80W. Now 60-160, it's progress. With 4th generation maybe they learn how to properly estimate power consumption.

BFL Problem: The board or something on it (but we don't think it's the ASIC chips) is drawing 2-3 times as much power as was expected, and we don't know what or why.  But we do know that the voltage regulator can't keep up.

BFL Solution: Add another voltage regulator so that together they'll be more able to handle the increased draw.

[johnyj]

And that 120W draw at run time is only when the chips running at 34Gh/s, so 240W draw for 68Gh/s  

[greyhawk]

"We are still working on the communication protocol"

[techwtf]

Anyone can tell me:
1 why only 15 nonce like response is returned?
2 what was he tested? single chip? hash core?
3 why it take ~20s to have the power to return idle?

1. sounds one one of the cores isn't working.
2. single chip on a board - each chip has 16 cores (see image)
3. presumably the test sends a block header (or equivalent) to each core and gets them to iterate over the nonce until the hash < difficulty - they program difficulty in - and the fact that they all came up with the same nonce means they are hashing.  It's the smallest unit of work that an ASIC that does SHA1-256(SHA1-256(data)) would have to do to prove it works.

Basically the video proves that chip works on a board, but only 15 of the 16 cores are hashing, and it uses a lot more power than expected (approx 4x)

Will

about 2:
One should be shocked if only a single chip is tested.
assume all 16 cores are working, 4GHash (per nonce) * 16 (cores) / 20 (second) = 3.276GH/s per chip, or 205MH/s per hash core.
with this watts shown, you can calculate the efficiency of MHash / Joule by yourself.

the machine should return 8 * 16 = 128 nonces (so the watts can be compared with avalon / asicminer), but why there is only 15...

[johnyj]

Anyone can tell me:
1 why only 15 nonce like response is returned?
2 what was he tested? single chip? hash core?
3 why it take ~20s to have the power to return idle?

1. sounds one one of the cores isn't working.
2. single chip on a board - each chip has 16 cores (see image)
3. presumably the test sends a block header (or equivalent) to each core and gets them to iterate over the nonce until the hash < difficulty - they program difficulty in - and the fact that they all came up with the same nonce means they are hashing.  It's the smallest unit of work that an ASIC that does SHA1-256(SHA1-256(data)) would have to do to prove it works.

Basically the video proves that chip works on a board, but only 15 of the 16 cores are hashing, and it uses a lot more power than expected (approx 4x)

Will

about 2:
One should be shocked if only a single chip is tested.
assume all 16 cores are working, 4GHash (per nonce) * 16 (cores) / 20 (second) = 3.276GH/s per chip, or 205MH/s per hash core.
with this watts shown, you can calculate the efficiency of MHash / Joule by yourself.

the machine should return 8 * 16 = 128 nonces (so the watts can be compared with avalon / asicminer), but why there is only 15...

He mentioned 16 Engines, not 16 cores, some kind of internal function that conbine 8 cores (half of one chip) to make one engine?

All the hashing engines return the same result, so I suppose that the slowest (most unlucky) engine need 18 seconds to finish the hashing function, average could be much less, say 1-2 seconds

[Tehfiend]

Yay! they are not a scam at least. This means they'll eventually ship a product.

fyi: we reached that stage 1.2 month before we shipped. Let's see how they handle their remainder issues.

The primeAsic scammers just released a video that shows just as much proof IMO

[BFL-Engineer]

Anyone can tell me:
1 why only 15 nonce like response is returned?
2 what was he tested? single chip? hash core?
3 why it take ~20s to have the power to return idle?

1) 15 nonce are returned because we run each engine the full range of 0 to 0xFFFFFFFF. They all find the same result, and we read all the engines when they're done.
2) 6 chips on board were tested with the same data. What was shown was the dump of 15 engines on one of the chips.
3) Because the processing range is not distributed among engines, so all the engines go from 0 to 0xFFFFFFFF.



Regards,
Nasser

[Frankie Delaney]

Anyone can tell me:
1 why only 15 nonce like response is returned?
2 what was he tested? single chip? hash core?
3 why it take ~20s to have the power to return idle?

1) 15 nonce are returned because we run each engine the full range of 0 to 0xFFFFFFFF. They all find the same result, and we read all the engines when they're done.
2) 6 chips on board were tested with the same data. What was shown was the dump of 15 engines on one of the chips.
3) Because the processing range is not distributed among engines, so all the engines go from 0 to 0xFFFFFFFF.



Regards,
Nasser

so the test was 15 of 16 engines on 6 of 8 chips? was this done for yeilds or something?

[Frizz23]

What is the die size again?

Package size is 11mm^2.

11mm^2

Die size is 7.5mm^2 if I remember correctly.

20W for 7.5mm^2 is quite a lot of heat to dissipate.

[k9quaint]

Well this is a small step in the right direction.
Maybe after 15 or 20 videos showing steady progress, I might believe they will ship something able to hash at some point in the future.

$30,000 dollars spent on a mini-rig last summer would have bought 5000 bitcoins. Those 5000 BTC would be worth $455,000 today.
BFL has some catching up to do.

[Frankie Delaney]

Well this is a small step in the right direction.
Maybe after 15 or 20 videos showing steady progress, I might believe they will ship something able to hash at some point in the future.

$30,000 dollars spent on a mini-rig last summer would have bought 5000 bitcoins. Those 5000 BTC would be worth $455,000 today.
BFL has some catching up to do.

I don't remember where the list is, but some guy bought 4 minirigs on june 23rd, when the price would have been $6.40/BTC. That $120,000 invested in bitcoins instead of butterflies would be worth $1.71 million dollars. Or alternatively, it would have bought 185 4 module avalons good for 15.8TH/s, as opposed to the 6TH/s he might get if BFL ever ships something.

[SebastianJu]

Well this is a small step in the right direction.
Maybe after 15 or 20 videos showing steady progress, I might believe they will ship something able to hash at some point in the future.

$30,000 dollars spent on a mini-rig last summer would have bought 5000 bitcoins. Those 5000 BTC would be worth $455,000 today.
BFL has some catching up to do.

I don't remember where the list is, but some guy bought 4 minirigs on june 23rd, when the price would have been $6.40/BTC. That $120,000 invested in bitcoins instead of butterflies would be worth $1.71 million dollars. Or alternatively, it would have bought 185 4 module avalons good for 15.8TH/s, as opposed to the 6TH/s he might get if BFL ever ships something.

Lets hope he had way more than that. Otherwise it would be really a pity.

[PuertoLibre]

Great. They are now in the prototype stage.

LOL

The guy in the video sounds like an apologetic scam artist....where do I know his voice from?

42 at idle and 160watt at full crank? Sounds like they missed their power targets if you ask me....

Just like the FPGA launch....

I don't remember hearing the guy say how many GH/s it was doing.

[PuertoLibre]

42W at idle?  Good lawd.  My x6500 FPGAs only burns up 17.2W at full load...

Didn't they like...put a dummy load in place of the ASIC chips and actually test these boards or something while they were waiting on chip delivery?

I guess not.

It was the engineering teams fault for not thinking of it?

[johnyj]

Well this is a small step in the right direction.
Maybe after 15 or 20 videos showing steady progress, I might believe they will ship something able to hash at some point in the future.

$30,000 dollars spent on a mini-rig last summer would have bought 5000 bitcoins. Those 5000 BTC would be worth $455,000 today.
BFL has some catching up to do.

I don't remember where the list is, but some guy bought 4 minirigs on june 23rd, when the price would have been $6.40/BTC. That $120,000 invested in bitcoins instead of butterflies would be worth $1.71 million dollars. Or alternatively, it would have bought 185 4 module avalons good for 15.8TH/s, as opposed to the 6TH/s he might get if BFL ever ships something.

Time travel trading is not relevant, if bitcoin today worth only $2 due to some severe bug in software, that guy can celebrate that he made the right decision to spend the coin before the price crashed

[BitcoinINV]

Great. They are now in the prototype stage.

LOL

They guy in the video sounds like an apologetic scam artist....where do I know his voice from?

42 at idle and 160watt at full crank? Sounds like they missed their power targets if you ask me....

Just like the FPGA launch....

I don't remember hearing the guy say how many GH/s it was doing.

No he just said they were doing what they were supposed to.... I guess they were all just supposed turn on lmfao?

[Dargo]

Well this is a small step in the right direction.
Maybe after 15 or 20 videos showing steady progress, I might believe they will ship something able to hash at some point in the future.

$30,000 dollars spent on a mini-rig last summer would have bought 5000 bitcoins. Those 5000 BTC would be worth $455,000 today.
BFL has some catching up to do.

I don't remember where the list is, but some guy bought 4 minirigs on june 23rd, when the price would have been $6.40/BTC. That $120,000 invested in bitcoins instead of butterflies would be worth $1.71 million dollars. Or alternatively, it would have bought 185 4 module avalons good for 15.8TH/s, as opposed to the 6TH/s he might get if BFL ever ships something.

https://bitcointalk.org/index.php?topic=89685.0

gigavps bought 4 minirigs. My memory is that bitlane also bought 4, but don't see him on the list now, so maybe he cancelled.

[willphase]

https://bitcointalk.org/index.php?topic=89685.0

gigavps bought 4 minirigs. My memory is that bitlane also bought 4, but don't see him on the list now, so maybe he cancelled.

gigavps probably already had (at least) four FPGA minirigs, so deciding to upgrade at the time was a no-brainer since he could get full price on each FPGA rig off each ASIC rig - and incidentally, mining on four minirigs in that time has probably made him around 10,000 BTC so I'm sure he's not complaining too hard.

Will

[PuertoLibre]

HAHAHAHAHAHAHAHAHAHAHAHAHA *dies*

So they've tested that the board can handle the data. They are still to:

Actually have a chip
Actually make the chip work
Actually bin the chip
Sort out the communication protocols, never mind proper USB interfaces etc
Get the power consumption down to ANYTHING near what they promised
Actually make it hash
Actually make it hash to the required speed
Actually make it hash consistently
Actually make it hashable 24/7
Actually prevent it overheating [That's 160W TDP on the equivalent of a dual slot GPU, which would typically warrant ~40CFM - they have 0, AND will be placing many in a an enclosed case]
Work out how to do the power distribution
Make the shit fit in the cases.
Work out sufficient testing algorithms and the logistics of burning in '300 units a day'
Sort out the logistics of labelling, warranties, who gets what.
Actually deliver anything that works for more than a few days.



ETA 5 months, GG.

The red part...

That is what I thought about when Inaba/Josh used to gloat about it's size. The funny pics of the FPGA were interesting with the fans beneath...

But on another point...anyone noticed that the heat sink was seemingly the aluminum one (for Mini-Single) and not the other variety of copper type the Single SC ASIC has.

I played this twice and I noticed that the person (Josh?) behind the camera fails to mention which rig this is. From the aluminum I am going to guess it is a Mini-Single @ 160watts.

Add a PC and you've got 300 watts. Does this imply the Single will be twice as hot with it's copper heat-sink? Now we have a clue as to the reason for the silence.....

------------------

As for the blue...I think the Mini-rigs are going to need quite a few power outlets if that is the case.

[PuertoLibre]

Im confused....
Now, Im not that versed in power circuitry ... but how the hell can the PCB power circuitry design be so shitty that the board itself is using nearly 2x the power than the chip? iirc the SC was supposed to be like 60watts right? The whole unit is using 150ish .. how the fuck can the board be using 90watts?!?!?
Did some noob engineer design the board or some shit? I mean really.

April fools?

[k9quaint]

Well this is a small step in the right direction.
Maybe after 15 or 20 videos showing steady progress, I might believe they will ship something able to hash at some point in the future.

$30,000 dollars spent on a mini-rig last summer would have bought 5000 bitcoins. Those 5000 BTC would be worth $455,000 today.
BFL has some catching up to do.

I don't remember where the list is, but some guy bought 4 minirigs on june 23rd, when the price would have been $6.40/BTC. That $120,000 invested in bitcoins instead of butterflies would be worth $1.71 million dollars. Or alternatively, it would have bought 185 4 module avalons good for 15.8TH/s, as opposed to the 6TH/s he might get if BFL ever ships something.

Time travel trading is not relevant, if bitcoin today worth only $2 due to some severe bug in software, that guy can celebrate that he made the right decision to spend the coin before the price crashed

This is not time travel trading. This is examining people who were trying to bet on BTC (who instead bet on magical thinking). I was pointing out that there was a very simple way to bet on BTC available to them at the time. Had they taken that simple bet, they would be millionaires.

[PuertoLibre]

Yay! they are not a scam at least. This means they'll eventually ship a product.

fyi: we reached that stage 1.2 month before we shipped. Let's see how they handle their remainder issues.

Someday you should let your customer know just how the development of the Avalon device had to go through to get to this point. I wouldn't mind hearing the backstory of how it came together.

[repentance]

The only party surprised by higher power usage appears to be BFL itself.. even after they failed to do correct estimates for their FPGAs.

<insert history repetition quote>

The "surprise" seems to be coming from the fact that it's only happening with one wafer - the one which was packaged.  It hasn't occurred with the tests they ran on the raw chips.  The question is whether wafers 3-6 will have the same problem - if they do, exactly why the packaged chips are having the problem and the raw chips aren't needs to be identified.

I vaguely recall Yifu predicting early this year that BFL would encounter this issue and I'm curious about why he thought this was the case. Also, were those who commented on the images of the packaged chips and said that the fill job was crappy seeing something which could be contributing to this issue?

That BFL is having issues shouldn't be surprising - the point of testing is to discover and resolve issues.  The problem stems more from them having consistently underestimated the time it will take them to solve any given issue.  I doubt they can realistically estimate when this one might be resolved and when "revised" units will be available - and that's a critical question because the bulk of their orders won't be filled from Batch 1 and 2.

[Tamerz]

I vaguely recall Yifu predicting early this year that BFL would encounter this issue and I'm curious about why he thought this was the case.

I remember this as well. Like they ran into the same thing early on.

[Dargo]

https://bitcointalk.org/index.php?topic=89685.0

gigavps bought 4 minirigs. My memory is that bitlane also bought 4, but don't see him on the list now, so maybe he cancelled.

gigavps probably already had (at least) four FPGA minirigs, so deciding to upgrade at the time was a no-brainer since he could get full price on each FPGA rig off each ASIC rig - and incidentally, mining on four minirigs in that time has probably made him around 10,000 BTC so I'm sure he's not complaining too hard.

Will

I agree - didn't intend to suggest gigvps was hurting, just that he is the only 4 minirig purchase according to the list.

[marra]

so this proves mini rigs are still a fiction... at 5KW it can't fit in regular power supply switch, at least not in europe, 3.5 KW max per fuse...

[beekeeper]

[BitSyncom]

I vaguely recall Yifu predicting early this year that BFL would encounter this issue and I'm curious about why he thought this was the case.

I remember this as well. Like they ran into the same thing early on.

not really, what I said was we also ran simulations for 65nm, and based on our results we knew they will run into this problem, and some other ones I shall not mention.

[itcamefrommars]

Give BFL a break everyone!  They're still waiting on their batch 2 avalon to arrive so that they can open it up and see how it's really done!
Now you can blame avalon for bfl's delay.  That and radio shack not having better than 7xxxseries reg's.

What really bugs me is the parted together feeling of generic USB hubs in daisy chained in the minirig (or what ever it's called... I'm starting to unconsciously block out anything BFL from my brain)  That's a dirt cheap and simple circuit to make well under the cost of purchased ones even at alibaba prices or china street prices ... or was that when they were just showing the prototype of the case with nothing of value.  At least the minirig doesn't look like a spray painted AppleTV or Macmini.

Avalon... when are you going to help these poor guys out ?  lol... you might save a lot of bfl customers from suicide/divorce/PTSD/unpaid mortgages/unpaid student loans/second jobs/etc... it'd be like mining for karma 

 

[witherworth]

Well this is a small step in the right direction.
Maybe after 15 or 20 videos showing steady progress, I might believe they will ship something able to hash at some point in the future.

$30,000 dollars spent on a mini-rig last summer would have bought 5000 bitcoins. Those 5000 BTC would be worth $455,000 today.
BFL has some catching up to do.

I don't remember where the list is, but some guy bought 4 minirigs on june 23rd, when the price would have been $6.40/BTC. That $120,000 invested in bitcoins instead of butterflies would be worth $1.71 million dollars. Or alternatively, it would have bought 185 4 module avalons good for 15.8TH/s, as opposed to the 6TH/s he might get if BFL ever ships something.

Time travel trading is not relevant, if bitcoin today worth only $2 due to some severe bug in software, that guy can celebrate that he made the right decision to spend the coin before the price crashed

This is not time travel trading. This is examining people who were trying to bet on BTC (who instead bet on magical thinking). I was pointing out that there was a very simple way to bet on BTC available to them at the time. Had they taken that simple bet, they would be millionaires.

Except you've got to take into account what a person would actually do. For myself, I wouldn't have invested in BTC itself, expecting the price to rise (I've actually been expecting it to fall). I did invest in hardware, because I know that I'll get returns on it eventually (even if it did take many months), and I felt it was a safer bet than buying BTC upright. Therefore, had I not invested in BFL hardware, I wouldn't have invested in BTC, and therefore didn't lose out on anything.

Arguing that I still have some lost profit would be like me arguing that the dollar you could have given to that homeless guy you passed is a lost investment because you didn't realize that he used that dollar to buy a lottery ticket and felt obligated to thank you back for your kindness by giving you thousands of dollars in return. It just doesn't make sense.

[Entropy-uc]

I managed to track down the chart I was talking about above.

My calculations put their reported power density right in the middle of the Pentium 4 space.

BFL is going to have to completely re-engineer their thermal solution if they can't get the power reduced by at least 30%.

[SebastianJu]

I managed to track down the chart I was talking about above.

My calculations put their reported power density right in the middle of the Pentium 4 space.

BFL is going to have to completely re-engineer their thermal solution if they can't get the power reduced by at least 30%.

Im not sure but isnt the graph misleading? I mean you only use the chip for w/cm² but one should take the fuel rod of a nuclear reactor only then too, isnt it? And then this value cant be true anymore am i wrong? It doesnt look correct to me that the power density in a chip can be near the power density created by a nuclear reaction. But im no physicist.

[ShadesOfMarble]

"It doesnt look correct to me" -> logarithmic scale

[SebastianJu]

"It doesnt look correct to me" -> logarithmic scale

Yes, but even then... the scale shows something like 90 W/cm² for a pentium 4 and maybe 250 W/cm² for a nuclear reactor? Thats the whole building then isnt it?

[Ekaros]

"It doesnt look correct to me" -> logarithmic scale

Yes, but even then... the scale shows something like 90 W/cm² for a pentium 4 and maybe 250 W/cm² for a nuclear reactor? Thats the whole building then isnt it?

90W under load seems to be true.

There is reason why nuclear rods are submerged in water, other than radiation. You really have to consider size difference. Pentium 4 chips were around 2cm^2 where as nuclear rods are much larger.

[jedunnigan]

Entropy can you be a bit more specific? What exactly would happen? Before a fire, that is...

[johnyj]

I guess in this case, even the VRM need to be actively cooled

[tvbcof]

...
Im not sure but isnt the graph misleading? I mean you only use the chip for w/cm² but one should take the fuel rod of a nuclear reactor only then too, isnt it? And then this value cant be true anymore am i wrong? It doesnt look correct to me that the power density in a chip can be near the power density created by a nuclear reaction. But im no physicist.

Seems intuitively OK to me.  The reaction in a nuclear reactor is a controlled criticality and needs to be within a manageable density range for long duration use.  It is also large in size so the total power output is quite large.  A rocket engine only runs briefly and consumes a huge amount of high energy fuel per unit time.

It would be interesting to know the power density within a fusion device in the brief moment the pit goes critical.

[superfastkyle]

So what you really mean is primeAsic now has the same amount of proof that Avalon does, lol

The primeAsic scammers just released a video that shows just as much proof IMO

[Entropy-uc]

Entropy can you be a bit more specific? What exactly would happen? Before a fire, that is...

The first thing is a huge heat sink and a really loud fan.  That's what we had on p4.

If your processor runs too hot, you get diffusion in the various materials that make up the device - this dramatically reduces the working life of the processor due to a range of failure modes.  Ask the GPU mining guys how fast that can happen.   

In the shorter term resistances rise with temperature, driving power demand up in a negative feedback loop.  I would guess that test video was just about the amount of time they could run chips without having the magic smoke come out of some components.

I don't know what I would do in BFL's shoes.  There isn't enough data to really understand what the situation is.  But what they reported is not good in the short term.

[SLok]

Code:

bfgminer version 2.99.2 - Started: [2013-03-30 23:37:17] - [  0 days 00:01:57
------------------------------------------------------------------------------
 5s:17.85 avg:16.58 u:14.26 Gh/s | A:342 R:0 S:0 HW:0 U:179.6/m
 ST: 2  DW: 10  GW: 3  LW: 494  GF: 0  NB: 1  AS: 0  RF: 0  E: 2.08
 Connected to 192.168.1.1 diff 8 with stratum as user test
 Block: ...ee2e8a72 #3123  Diff:153  Started: [23:37:17]  Best share: 528
------------------------------------------------------------------------------
 [P]ool management [S]ettings [D]isplay options [Q]uit
 BFL 0:  45.0C         | 17.83/17.52/14.58Gh/s | A:343 R:0 HW:0 U:180.16/m
------------------------------------------------------------------------------

 [2013-03-30 23:38:52] Stratum from pool 0 requested work update
 [2013-03-30 23:38:52] Accepted 0ecf365f BFL 0  Diff 17/1
 [2013-03-30 23:38:53] Accepted a91d942c BFL 0  Diff 1/1
 [2013-03-30 23:38:53] Accepted 15d55ee2 BFL 0  Diff 11/1
 [2013-03-30 23:38:53] Accepted be2972c6 BFL 0  Diff 1/1
 [2013-03-30 23:38:53] Accepted 9e87d2fd BFL 0  Diff 1/1
 [2013-03-30 23:38:53] Accepted 7f854200 BFL 0  Diff 2/1
 [2013-03-30 23:38:54] Accepted 03c5d641 BFL 0  Diff 67/8
 [2013-03-30 23:39:08] Accepted 040aab96 BFL 0  Diff 63/8
 [2013-03-30 23:39:09] Accepted 14997fe3 BFL 0  Diff 12/8
 [2013-03-30 23:39:10] Accepted 123ded4b BFL 0  Diff 14/8
 [2013-03-30 23:39:11] Accepted 0453e37f BFL 0  Diff 59/8
 [2013-03-30 23:39:14] Accepted 06bd2321 BFL 0  Diff 37/8

https://bitcointalk.org/index.php?topic=78192.msg1707177#msg1707177

[k9quaint]

Entropy can you be a bit more specific? What exactly would happen? Before a fire, that is...

The first thing is a huge heat sink and a really loud fan.  That's what we had on p4.

If your processor runs too hot, you get diffusion in the various materials that make up the device - this dramatically reduces the working life of the processor due to a range of failure modes.  Ask the GPU mining guys how fast that can happen.   

In the shorter term resistances rise with temperature, driving power demand up in a negative feedback loop.  I would guess that test video was just about the amount of time they could run chips without having the magic smoke come out of some components.

I don't know what I would do in BFL's shoes.  There isn't enough data to really understand what the situation is.  But what they reported is not good in the short term.

Isn't the usual response to too much heat to lower the clock speed and/or lower the voltage?
The unusual response being custom machined copper water blocks with liquid cooling.

[atomicdog]

after watching the video i don't know how they could give an estimate of shipping in a week given what the video shows.  there are a myriad of questions that the video brings up.  i don't see how the solution could only be adding additional voltage regulators to a new pcb.  do they even know if they will need a larger heatsink/fan and will it even fit in their nifty case? wow what a butterfluck...

[tvbcof]

...
It would be interesting to know the power density within a fusion device in the brief moment the pit goes critical.

that would be such a big number it's hard to think about.  And so classified I'm a little nervous even replying to you.

Were only talking order-of-magnitude estimates, and the megaton yield and rough dimensions of are reasonably well know.  But not by me, however, and I don't feel like looking it up or embarrassing myself with the mathematics which are rusty.  Suffice it to say, though, that 'big' is probably descriptive enough

[antirack]

Bottom line:  Intel P4 processors were the highest power density systems we shipped.  Once you get in that range, really bad things happen.

I believe IBMs Power PC 4 CPU had some hot spots of up to 140 W/cm2. Apple's G5 had two dies under the hood with a total flux of around 200 W/cm2. I think your "really bad things happen" is spot on, Apple had a leaking water cooling system

[beekeeper]

40W per square centimeter is way too much. Put your palm around a light bulb. Then imagine all that heat burning your hand comes trough a finger nail, which would be the chip..

As someone said:

The more important question is:
WILL IT BLEND?

or, "will it melt?"

Ofc, it is possible to design packages and cooling systems to use this chip, but I doubt it can be done safely in a week.

[PulsedMedia]

guys guys, the chip power usage is not the full 195W, it's the whole board.
The board itself will consume a lot of the power, most in this case due to going badly out of spec.

If the power reqs go far beyond their nominal range efficiency goes shit, requiring higher amperage for the specific output.
Power regs ran beyond their spec can consume insane amounts of power, which mostly goes in to heating.
So if it was rated for 5A (60W), but the board required 7.5A (90W), due to efficiency going shit it would jump straight to 9A, and further due to heat above 10A. Then the traces might not be large enough for the reg, further putting demand for higher current due to increased resistance. Voltage also drops, it might reach a point where voltage drops, and again demand for current goes higher. To get 20A @ 3.3V with 11V takes a lot more current than at 12V. At 11V it's 6A, at 12V it's 5.5A.
Current is what causes things to heat up and is most affected by resistance. You can probably transmit 1kW @ 1000V over the same traces which can barely handle 50W @ 3.3V. (Given insulation is sufficient)

All those figures are just speculative, depends on the regs used, trace sizes etc. Just basics really, but you get the idea.

So it's entirely possible adding another reg there will fix it.

I'm not an electronics engineer, but i've made my dabble into electronics

[wndrbr3d]

guys guys, the chip power usage is not the full 195W, it's the whole board.
The board itself will consume a lot of the power, most in this case due to going badly out of spec.

I'm no physics major, but my pappy told me long ago that heat is the result of wasted energy. So I don't think it matters WHERE the energy is being spent, odds are it'll generate a fair amount of heat.

[Meizirkki]

guys guys, the chip power usage is not the full 195W, it's the whole board.
The board itself will consume a lot of the power, most in this case due to going badly out of spec.

There's no way they designed the board for 60W total. Regulators aren't going to do twice their designed output.

[Frizz23]

[Bogart]

it's entirely possible adding another reg there will fix it.

Surely there's a reason the boards didn't run as intended in the first place.

I don't see how adding another voltage regulator will magically fix it.  It might improve efficiency somewhat, but that's hardly a complete fix.