Author

Topic: Who will be the next "bASIC"? (Read 948 times)

hero member
Activity: 784
Merit: 1004
Glow Stick Dance!
November 05, 2013, 01:55:30 AM
#19
cointerra has been quiet lately it seems

Concur.  I see them as being the most likely, followed by HashFast.

The idea of 250W TDP in a single die just seems like the wrong approach to me.
Esp when that's the theoretical power consumption - which no one apart from Avalon even got close to. Estimate 250 = spec for 450, which isn't going to happen.

Well KNC got it right (came in significantly under simulation) so it is more than just Avalon.

Hyptohetically there is a miss and it does use 450W vs 250W at nominal voltage and clock.  Power is all relative.  It would be 450W at a given clockrate and voltage not at all voltage and clock speeds.  Lower the voltage and power drops by the square.  Lower the clock and power drops linearly.  If the system absolutely can't use more than 250W there is some combination of clockrate and voltage that gets you there.

This is exactly what Bitfury did.  
Of course this is true, but only if we don't care about hash rate promises. Which we do.

Well once again it is no different than if a small chip misses.  Bitfury did and used more chips to deliver the same hashpower.  Obviously they would have rather not missed and made an even larger profit but they adapted and met their promises.

If you miss you need to clock/volt down the chip and use more chips. 500 GH/s chip or 5 GH/s chip you still need to adjust.

... or they could waste precious time and do a respin.
donator
Activity: 1218
Merit: 1079
Gerald Davis
November 05, 2013, 12:35:34 AM
#18
cointerra has been quiet lately it seems

Concur.  I see them as being the most likely, followed by HashFast.

The idea of 250W TDP in a single die just seems like the wrong approach to me.
Esp when that's the theoretical power consumption - which no one apart from Avalon even got close to. Estimate 250 = spec for 450, which isn't going to happen.

Well KNC got it right (came in significantly under simulation) so it is more than just Avalon.

Hyptohetically there is a miss and it does use 450W vs 250W at nominal voltage and clock.  Power is all relative.  It would be 450W at a given clockrate and voltage not at all voltage and clock speeds.  Lower the voltage and power drops by the square.  Lower the clock and power drops linearly.  If the system absolutely can't use more than 250W there is some combination of clockrate and voltage that gets you there.

This is exactly what Bitfury did.  
Of course this is true, but only if we don't care about hash rate promises. Which we do.

Well once again it is no different than if a small chip misses.  Bitfury did and used more chips to deliver the same hashpower.  Obviously they would have rather not missed and made an even larger profit but they adapted and met their promises.

If you miss you need to clock/volt down the chip and use more chips. 500 GH/s chip or 5 GH/s chip you still need to adjust.
legendary
Activity: 1666
Merit: 1183
dogiecoin.com
November 05, 2013, 12:19:07 AM
#17
cointerra has been quiet lately it seems

Concur.  I see them as being the most likely, followed by HashFast.

The idea of 250W TDP in a single die just seems like the wrong approach to me.
Esp when that's the theoretical power consumption - which no one apart from Avalon even got close to. Estimate 250 = spec for 450, which isn't going to happen.

Well KNC got it right (came in significantly under simulation) so it is more than just Avalon.

Hyptohetically there is a miss and it does use 450W vs 250W at nominal voltage and clock.  Power is all relative.  It would be 450W at a given clockrate and voltage not at all voltage and clock speeds.  Lower the voltage and power drops by the square.  Lower the clock and power drops linearly.  If the system absolutely can't use more than 250W there is some combination of clockrate and voltage that gets you there.

This is exactly what Bitfury did. 
Of course this is true, but only if we don't care about hash rate promises. Which we do.
donator
Activity: 1218
Merit: 1079
Gerald Davis
November 05, 2013, 12:02:50 AM
#16
cointerra has been quiet lately it seems

Concur.  I see them as being the most likely, followed by HashFast.

The idea of 250W TDP in a single die just seems like the wrong approach to me.
Esp when that's the theoretical power consumption - which no one apart from Avalon even got close to. Estimate 250 = spec for 450, which isn't going to happen.

Well KNC got it right (came in significantly under simulation) so it is more than just Avalon.

Hyptohetically there is a miss and it does use 450W vs 250W at nominal voltage and clock.  Power is all relative.  It would be 450W at a given clockrate and voltage not at all voltage and clock speeds.  Lower the voltage and power drops by the square.  Lower the clock and power drops linearly.  If the system absolutely can't use more than 250W there is some combination of clockrate and voltage that gets you there.

This is exactly what Bitfury did. 
donator
Activity: 1218
Merit: 1079
Gerald Davis
November 05, 2013, 12:00:11 AM
#15
Too soon for 14nm then.  22nm?   Grin

About 2-3 years too son for economical 20/22nm.
legendary
Activity: 1666
Merit: 1183
dogiecoin.com
November 04, 2013, 11:53:48 PM
#14
cointerra has been quiet lately it seems

Concur.  I see them as being the most likely, followed by HashFast.

The idea of 250W TDP in a single die just seems like the wrong approach to me.
Esp when that's the theoretical power consumption - which no one apart from Avalon even got close to. Estimate 250 = spec for 450, which isn't going to happen.
legendary
Activity: 966
Merit: 1000
November 04, 2013, 10:29:03 PM
#13
cointerra has been quiet lately it seems

Concur.  I see them as being the most likely, followed by HashFast.

The idea of 250W TDP in a single die just seems like the wrong approach to me.
legendary
Activity: 2114
Merit: 1005
ASIC Wannabe
November 04, 2013, 08:35:20 PM
#12
28 nm as well


Well, I hope that KNC has room for improvement on their first pass at 28nm.

I'd really be interested to see what Bitfury can do at 28nm since they are already more efficient at 55nm than KNC is at 28nm.  And I think that Bitfury still has some improvements they can make to get the specs of their 55nm die closer to the projected 5GH/s per chip that they were aiming for.  They may even be able to go to a less costly die process like 32nm and still kick everyone's ass.

I love that guy.

I used to love him too, but not anymore because of his high prices.

in a few days all the october pre-orders will be fulfiled, and i expect a drop to reasonable prices ($12 per chip and $300-350 per h-card)
hero member
Activity: 742
Merit: 500
November 04, 2013, 07:58:54 PM
#11
Not BFL, since they just announced that the first Monarchs should be shipped by the end of December, with HashTrade, founded by the Jacobson Brothers, receiving their units in about March/April.

BTW, has anybody been able to hunt down the tx for that $1M down payment transaction?

We are talking about Bitcoin miners here, not tea cup warmers Tongue (even if it's a million USD worth of them)  Wink
legendary
Activity: 1918
Merit: 1570
Bitcoin: An Idea Worth Spending
November 04, 2013, 07:43:52 PM
#10
Not BFL, since they just announced that the first Monarchs should be shipped by the end of December, with HashTrade, founded by the Jacobson Brothers, receiving their units in about March/April.

BTW, has anybody been able to hunt down the tx for that $1M down payment transaction?
hero member
Activity: 784
Merit: 1004
Glow Stick Dance!
November 04, 2013, 07:32:05 PM
#9
28 nm as well


Well, I hope that KNC has room for improvement on their first pass at 28nm.

I'd really be interested to see what Bitfury can do at 28nm since they are already more efficient at 55nm than KNC is at 28nm.  And I think that Bitfury still has some improvements they can make to get the specs of their 55nm die closer to the projected 5GH/s per chip that they were aiming for.  They may even be able to go to a less costly die process like 32nm and still kick everyone's ass.

I love that guy.

I used to love him too, but not anymore because of his high prices.

Are we referring to "Tytus" when we say "Bitfury"?  He is kind of an enigma... unless you can speak Russian, I guess.  He never posts in the English language forums.  It would be nice to know what he was up to.
legendary
Activity: 1904
Merit: 1007
November 04, 2013, 07:26:21 PM
#8
28 nm as well


Well, I hope that KNC has room for improvement on their first pass at 28nm.

I'd really be interested to see what Bitfury can do at 28nm since they are already more efficient at 55nm than KNC is at 28nm.  And I think that Bitfury still has some improvements they can make to get the specs of their 55nm die closer to the projected 5GH/s per chip that they were aiming for.  They may even be able to go to a less costly die process like 32nm and still kick everyone's ass.

I love that guy.

I used to love him too, but not anymore because of his high prices.
legendary
Activity: 1512
Merit: 1000
November 04, 2013, 07:04:47 PM
#7
28 nm as well


Well, I hope that KNC has room for improvement on their first pass at 28nm.

I'd really be interested to see what Bitfury can do at 28nm since they are already more efficient at 55nm than KNC is at 28nm.  And I think that Bitfury still has some improvements they can make to get the specs of their 55nm die closer to the projected 5GH/s per chip that they were aiming for.  They may even be able to go to a less costly die process like 32nm and still kick everyone's ass.

I love that guy.
hero member
Activity: 784
Merit: 1004
Glow Stick Dance!
November 04, 2013, 06:58:32 PM
#6
28 nm as well


Well, I hope that KNC has room for improvement on their first pass at 28nm.

I'd really be interested to see what Bitfury can do at 28nm since they are already more efficient at 55nm than KNC is at 28nm.  And I think that Bitfury still has some improvements they can make to get the specs of their 55nm die closer to the projected 5GH/s per chip that they were aiming for.  They may even be able to go to a less costly die process like 32nm and still kick everyone's ass.
hero member
Activity: 742
Merit: 500
November 04, 2013, 06:38:51 PM
#5
28 nm as well
hero member
Activity: 784
Merit: 1004
Glow Stick Dance!
November 04, 2013, 06:37:52 PM
#4
KNC gen 2 in March 2014  Cheesy





(half trolling, would not be that surprised though...)



Too soon for 14nm then.  22nm?   Grin
hero member
Activity: 742
Merit: 500
November 04, 2013, 06:35:28 PM
#3
KNC gen 2 in March 2014  Cheesy





(half trolling, would not be that surprised though... and yes, gen 2 will likely be 28 nm as well and no, I don't want to defend that statement)

legendary
Activity: 2114
Merit: 1005
ASIC Wannabe
November 04, 2013, 06:29:22 PM
#2
Except for KNC, it seems like all of the other ASIC vendors are having difficulty bringing their 28nm designs to market.  I'm starting to think some of the millions of preorder dollars and bitcoins are in possible danger of never being returned if the designs ultimately fail.

And it doesn't help confidence levels when the ASIC vendors stop communicating their progress and future plans!  

So what's your guess on who will be the next bASIC?  Or do you think everyone will at least deliver some kind of working miner to investors whether it meets planned specs or not?

cointerra has been quiet lately it seems (but i dont actively follow them) but they were not planned to launch until 2014
hero member
Activity: 784
Merit: 1004
Glow Stick Dance!
November 04, 2013, 06:11:33 PM
#1
Except for KNC, it seems like all of the other ASIC vendors are having difficulty bringing their 28nm designs to market.  I'm starting to think some of the millions of preorder dollars and bitcoins are in possible danger of never being returned if the designs ultimately fail.

And it doesn't help confidence levels when the ASIC vendors stop communicating their progress and future plans!  

So what's your guess on who will be the next bASIC?  Or do you think everyone will at least deliver some kind of working miner to investors whether it meets planned specs or not?
Jump to: