Topic: Bitfury chip integration (Read 2330 times)
Probably www.bitfurystrikesback.com/
So you have to lay down a $1M order on the chips sight-unseen in order to get information about the chips? Either that's not correct or that's powerfully stupid.
"punin"/bitfury is saying that he will have smaller batches on "my site" without mentioning what "my site" is.
probably just bitfury's main site.
He also says that they currently have 98 chips (had 100, burned two by going to 180 Gh/s in immersion!).
Mass production-most likely late March-April (without overpay for a "hot lot").
Engineering schemes delivery in late Feb to those who pay now.
For 100+ GHs per chip you'll need a heat sink
From the info already in the open air I'm planning to do string design , but kind of a hybrid with DC/DC group just to play with the Voltage
There is no chip specs yet .
A lot of open questions like Chaining of SPC ...
I think we can trust that the chip is tailored for a string design.From the info already in the open air I'm planning to do string design , but kind of a hybrid with DC/DC group just to play with the Voltage
There is no chip specs yet .
A lot of open questions like Chaining of SPC ...
Are there no seperate VDDio and VSSio pins?
Weren't the BF3500 and BF4500 based on one of their 55nm designs?
Yeah, I think so. And there have been some "out in the wild"
Weren't the BF3500 and BF4500 based on one of their 55nm designs?
Look at previous boards and the board in the test - they're almost bare.
Are there reference designs or public information regarding any other than the 55nm bitfury generation? It would be a waste not to put a heatsink on that chip - or immersion cool it, like they do themselfs.
AFAIK no ones even benchmarked a 28nm chip in the wild, every machine must have been sold under a solid NDA or mandatory purchase order. Its pretty crazy if you think about it, 1000s and 1000s of BF3500s and BF4500s and not one has ever been 'tested'.
For 100+ GHs per chip you'll need a heat sink
From the info already in the open air I'm planning to do string design , but kind of a hybrid with DC/DC group just to play with the Voltage
There is no chip specs yet .
A lot of open questions like Chaining of SPC ...
I think we can trust that the chip is tailored for a string design.From the info already in the open air I'm planning to do string design , but kind of a hybrid with DC/DC group just to play with the Voltage
There is no chip specs yet .
A lot of open questions like Chaining of SPC ...
Are there no seperate VDDio and VSSio pins?
For 100+ GHs per chip you'll need a heat sink
From the info already in the open air I'm planning to do string design , but kind of a hybrid with DC/DC group just to play with the Voltage
There is no chip specs yet .
A lot of open questions like Chaining of SPC ...
From the info already in the open air I'm planning to do string design , but kind of a hybrid with DC/DC group just to play with the Voltage
There is no chip specs yet .
A lot of open questions like Chaining of SPC ...
Look at previous boards and the board in the test - they're almost bare.
Are there reference designs or public information regarding any other than the 55nm bitfury generation? It would be a waste not to put a heatsink on that chip - or immersion cool it, like they do themselfs.
I guess they will sell the chip with respect to the nominal speed of 100 GH/s for $13. But at this operation point you will not have 0.06 J/GH.
Yeah my interpretation is its more of a headline efficiency which could be used in the second half of this gen (like S4/S4+).Anyway, this would still be a good price.
$0.13/GH certainly isn't terrible when you consider how cheap their reference design is to make and should allow the OEMs to get by at least for now.Have you got information regarding the reference design or are you just guessing?
Look at previous boards and the board in the test - they're almost bare.
I guess they will sell the chip with respect to the nominal speed of 100 GH/s for $13. But at this operation point you will not have 0.06 J/GH.
Yeah my interpretation is its more of a headline efficiency which could be used in the second half of this gen (like S4/S4+).Anyway, this would still be a good price.
$0.13/GH certainly isn't terrible when you consider how cheap their reference design is to make and should allow the OEMs to get by at least for now.Have you got information regarding the reference design or are you just guessing?
I guess they will sell the chip with respect to the nominal speed of 100 GH/s for $13. But at this operation point you will not have 0.06 J/GH.
Yeah my interpretation is its more of a headline efficiency which could be used in the second half of this gen (like S4/S4+).Anyway, this would still be a good price.
$0.13/GH certainly isn't terrible when you consider how cheap their reference design is to make and should allow the OEMs to get by at least for now. So 13 USD per chip for 50 GHs, way to expensive for production
For engineering it is OK
For engineering it is OK
That's the price of the efficiency! The chip performance is low.
I guess they will sell the chip with respect to the nominal speed of 100 GH/s for $13. But at this operation point you will not have 0.06 J/GH.
According to their press release it should be approximately 0.12 J/GH instead.
Anyway, this would still be a good price.
So air about 140-150 tops with heat sinks and stuff
So 13 USD per chip for 50 GHs, way to expensive for production
For engineering it is OK
For engineering it is OK
oтнимaeм пoлгoдa oт cтapтa пpoдaж и пoлyчaeм вpeмя кoгдa эти дeвaйcы yжe вoвcю xyяpили в дц нa вaлepy
Heт, в этoт paз мы пocкpoмнee 
Пpoдaём микpocxeмы пpямo ceйчac. У нac y caмиx иx тoлькo пoкa Cpoки дocтaвoк y нac гдe-тo нa втopyю пoлoвинy мapтa - нaчaлo aпpeля (ecли нe дoплaтить фaбpикe зa XoтЛoт), инжeнepкy пocтaвим к кoнцy фeвpaля тeм ктo ceйчac зaкaзывaeт. Mинимaльный зaкaз нa $1M (ecли пpямo oт нac), или пoдoждaть пoкa нe выcтaвлю в мaлeнькиx пapтияx нa мoём caйтe.
Meaning: They pushed the chip to 180GH/s before it died... Immersion cooling of course.
To my knowledge BitFury has remained pretty silent on this chip as far as development and integration goes. As with basically every other BitFury announcement we should probably assume this chip doesn't exist until we see concrete proof of it in action.
https://youtu.be/3zPpj1JYw38
Cool! Thanks for showing me this video
So 13 USD per chip for 50 GHs, way to expensive for production
For engineering it is OK
For engineering it is OK
And those who have looked at them an know a lot like Dogie mentioned are under NDA.
[I'm not under NDA with them at this point]
And those who have looked at them an know a lot like Dogie mentioned are under NDA.
[I'm not under NDA with them at this point]
NDA for pricing is OK
For technical docs no so much.
For technical docs no so much.
Jump to: