Pages:
Author

Topic: Cointerra AIRE Miner 16nm PreOrder - page 6. (Read 19889 times)

hero member
Activity: 702
Merit: 500
September 22, 2014, 09:50:07 AM
#84
most of the guys (knc, hashfast, cointerra etc) were taping out their 28nm with both tsmc and gf and receiving silicon back some 2 months later (thats after paying ridiculously high 'expedite fees'.  if you dont pay the expedite expect 3 months)  cointerra did it last time in 7 weeks.  it'll be longer than 28nm for 16nm cos its so new... but not much longer.  so lets say - for arguments' sake they tape out in October...  they would expect to see 1st silicon in early Jan (or late Dec if theyre really lucky), and volume production later in Q1 onwards.

How can they tapeout now when 16nm production is supposed to begin in q1 2015?

20nm production was said to begin in q1 2014 and we all know how that turned out for KNC.

when tsmc lets you tape out, thats when you get your production slot.  its possible theyre not doing volume production til q1 2015, but bitcoin asics are hardly volume

Great so this will be a test bed for their new 16nm production runs, the yields will be awful.

don't know.. a press release i read says its on the same line as their 20nm so they're expecting really good yields.

also, bitcoin chips are fantastic for yield because there's so much redundancy and repetition (repeated circuits) that they're pretty resilient to yield issues

edit:  here's the link to the press about yields etc...  http://blogs.barrons.com/asiastocks/2014/08/25/tsmc-to-start-16nm-production-earlier-heat-up-competition/
full member
Activity: 168
Merit: 100
September 22, 2014, 09:48:26 AM
#83
most of the guys (knc, hashfast, cointerra etc) were taping out their 28nm with both tsmc and gf and receiving silicon back some 2 months later (thats after paying ridiculously high 'expedite fees'.  if you dont pay the expedite expect 3 months)  cointerra did it last time in 7 weeks.  it'll be longer than 28nm for 16nm cos its so new... but not much longer.  so lets say - for arguments' sake they tape out in October...  they would expect to see 1st silicon in early Jan (or late Dec if theyre really lucky), and volume production later in Q1 onwards.

How can they tapeout now when 16nm production is supposed to begin in q1 2015?

20nm production was said to begin in q1 2014 and we all know how that turned out for KNC.

when tsmc lets you tape out, thats when you get your production slot.  its possible theyre not doing volume production til q1 2015, but bitcoin asics are hardly volume

Great so this will be a test bed for their new 16nm production runs, the yields will be awful.
hero member
Activity: 702
Merit: 500
September 22, 2014, 07:57:17 AM
#82
most of the guys (knc, hashfast, cointerra etc) were taping out their 28nm with both tsmc and gf and receiving silicon back some 2 months later (thats after paying ridiculously high 'expedite fees'.  if you dont pay the expedite expect 3 months)  cointerra did it last time in 7 weeks.  it'll be longer than 28nm for 16nm cos its so new... but not much longer.  so lets say - for arguments' sake they tape out in October...  they would expect to see 1st silicon in early Jan (or late Dec if theyre really lucky), and volume production later in Q1 onwards.

How can they tapeout now when 16nm production is supposed to begin in q1 2015?

20nm production was said to begin in q1 2014 and we all know how that turned out for KNC.

when tsmc lets you tape out, thats when you get your production slot.  its possible theyre not doing volume production til q1 2015, but bitcoin asics are hardly volume
legendary
Activity: 1736
Merit: 1001
September 22, 2014, 07:53:48 AM
#81
most of the guys (knc, hashfast, cointerra etc) were taping out their 28nm with both tsmc and gf and receiving silicon back some 2 months later (thats after paying ridiculously high 'expedite fees'.  if you dont pay the expedite expect 3 months)  cointerra did it last time in 7 weeks.  it'll be longer than 28nm for 16nm cos its so new... but not much longer.  so lets say - for arguments' sake they tape out in October...  they would expect to see 1st silicon in early Jan (or late Dec if theyre really lucky), and volume production later in Q1 onwards.

How can they tapeout now when 16nm production is supposed to begin in q1 2015?

20nm production was said to begin in q1 2014 and we all know how that turned out for KNC.

I doubt they will also they cant even get 20nm right.
hero member
Activity: 770
Merit: 509
September 22, 2014, 07:47:09 AM
#80
most of the guys (knc, hashfast, cointerra etc) were taping out their 28nm with both tsmc and gf and receiving silicon back some 2 months later (thats after paying ridiculously high 'expedite fees'.  if you dont pay the expedite expect 3 months)  cointerra did it last time in 7 weeks.  it'll be longer than 28nm for 16nm cos its so new... but not much longer.  so lets say - for arguments' sake they tape out in October...  they would expect to see 1st silicon in early Jan (or late Dec if theyre really lucky), and volume production later in Q1 onwards.

How can they tapeout now when 16nm production is supposed to begin in q1 2015?

20nm production was said to begin in q1 2014 and we all know how that turned out for KNC.
hero member
Activity: 702
Merit: 500
September 22, 2014, 07:42:05 AM
#79
Maybe someone should give them Spondoolies web address and suggest they buy some SP35's? Makes much more sense, dontcha think?
Changing mode from rage to speculation, it's fair to say that the way Guy dances around the discussion of the Cointerra announcement and "that sounds and looks familiar" comments that perhaps it already is spondoolies tech they plan to sell anyway. Either that or someone else's tech that Guy knows about. I don't think Cointerra will be doing anything in house this time around.

just to quell the speculation, Cointerra designed their new 16nm chip completely in-house in conjunction with tsmc and guc.  and its done, dusted, finished and ready for tape-out.  the power and timings they've quoted are from post layout simulations (the most accurate kind of simulation).  anyone who gives power numbers when they haven't completed physical design, is quoting theoretical numbers that must be taken with a pinch of salt.
Ready for Tape Out?Huh
Then they have big balls claiming you will get your order in the first quarter.   Can you walk us through the project time line on 16NM?   If tape out goes perfect and there is no one else in line, when would the masks be done?   Is it fair to say mid November at the earliest?   Then you kind of need to order wafers.
Geez.   Q1 is really aggressive, isn't it?   Or am I just overestimating the number of days this will take just to get wafers?

most of the guys (knc, hashfast, cointerra etc) were taping out their 28nm with both tsmc and gf and receiving silicon back some 2 months later (thats after paying ridiculously high 'expedite fees'.  if you dont pay the expedite expect 3 months)  cointerra did it last time in 7 weeks.  it'll be longer than 28nm for 16nm cos its so new... but not much longer.  so lets say - for arguments' sake they tape out in October...  they would expect to see 1st silicon in early Jan (or late Dec if theyre really lucky), and volume production later in Q1 onwards.
donator
Activity: 714
Merit: 510
Preaching the gospel of Satoshi
September 22, 2014, 01:08:31 AM
#78
Maybe someone should give them Spondoolies web address and suggest they buy some SP35's? Makes much more sense, dontcha think?
Changing mode from rage to speculation, it's fair to say that the way Guy dances around the discussion of the Cointerra announcement and "that sounds and looks familiar" comments that perhaps it already is spondoolies tech they plan to sell anyway. Either that or someone else's tech that Guy knows about. I don't think Cointerra will be doing anything in house this time around.

just to quell the speculation, Cointerra designed their new 16nm chip completely in-house in conjunction with tsmc and guc.  and its done, dusted, finished and ready for tape-out.  the power and timings they've quoted are from post layout simulations (the most accurate kind of simulation).  anyone who gives power numbers when they haven't completed physical design, is quoting theoretical numbers that must be taken with a pinch of salt.

Funny, TSMC is not ready for FinFet 16nm, they estimate to be production ready by Q1 2015 and their primary client is Apple and their A9 chips
If CoinTerra officially said that (citation?), sounds like extremely optimistic at best and scammy at worst.
legendary
Activity: 2478
Merit: 1020
Be A Digital Miner
September 21, 2014, 10:56:40 PM
#77
Maybe someone should give them Spondoolies web address and suggest they buy some SP35's? Makes much more sense, dontcha think?
Changing mode from rage to speculation, it's fair to say that the way Guy dances around the discussion of the Cointerra announcement and "that sounds and looks familiar" comments that perhaps it already is spondoolies tech they plan to sell anyway. Either that or someone else's tech that Guy knows about. I don't think Cointerra will be doing anything in house this time around.

just to quell the speculation, Cointerra designed their new 16nm chip completely in-house in conjunction with tsmc and guc.  and its done, dusted, finished and ready for tape-out.  the power and timings they've quoted are from post layout simulations (the most accurate kind of simulation).  anyone who gives power numbers when they haven't completed physical design, is quoting theoretical numbers that must be taken with a pinch of salt.
Ready for Tape Out?Huh
Then they have big balls claiming you will get your order in the first quarter.   Can you walk us through the project time line on 16NM?   If tape out goes perfect and there is no one else in line, when would the masks be done?   Is it fair to say mid November at the earliest?   Then you kind of need to order wafers.
Geez.   Q1 is really aggressive, isn't it?   Or am I just overestimating the number of days this will take just to get wafers?

legendary
Activity: 2128
Merit: 1073
September 21, 2014, 06:45:36 PM
#76
the power and timings they've quoted are from post layout simulations (the most accurate kind of simulation).
The "most accurate kind of simulation" is all-analog simulation using BSIM or equivalent. I doubt that they did that. As someone mentioned earlier their J/GH numbers are in the "standard cell" range, thus points to a design flow that is intentionally approximate.
hero member
Activity: 702
Merit: 500
September 21, 2014, 06:14:03 PM
#75
Maybe someone should give them Spondoolies web address and suggest they buy some SP35's? Makes much more sense, dontcha think?
Changing mode from rage to speculation, it's fair to say that the way Guy dances around the discussion of the Cointerra announcement and "that sounds and looks familiar" comments that perhaps it already is spondoolies tech they plan to sell anyway. Either that or someone else's tech that Guy knows about. I don't think Cointerra will be doing anything in house this time around.

just to quell the speculation, Cointerra designed their new 16nm chip completely in-house in conjunction with tsmc and guc.  and its done, dusted, finished and ready for tape-out.  the power and timings they've quoted are from post layout simulations (the most accurate kind of simulation).  anyone who gives power numbers when they haven't completed physical design, is quoting theoretical numbers that must be taken with a pinch of salt.
hero member
Activity: 572
Merit: 500
September 21, 2014, 05:37:50 PM
#74
If it's SP-Tech technology, it'll have drivers Smiley This would explain why cointerra-aire does not need one in particular Smiley However it needs a PI/Bone anyway outside the rack case, which is enough hassle if you host .. and since the software falls in the owner's task list, they don't even look to build anything for it. Very risky to pre-order hardware, not even knowing you can mine with it.
-ck
legendary
Activity: 4088
Merit: 1631
Ruu \o/
September 21, 2014, 05:19:08 PM
#73
Maybe someone should give them Spondoolies web address and suggest they buy some SP35's? Makes much more sense, dontcha think?
Changing mode from rage to speculation, it's fair to say that the way Guy dances around the discussion of the Cointerra announcement and "that sounds and looks familiar" comments that perhaps it already is spondoolies tech they plan to sell anyway. Either that or someone else's tech that Guy knows about. I don't think Cointerra will be doing anything in house this time around.
sr. member
Activity: 441
Merit: 250
September 21, 2014, 05:15:27 PM
#72
And where did they get (or plan to get) the $7 million of NRE from?
They are hoping that the bitcoin community will give them a free loan... aka pre-orders.

So assuming a margin of 40% they'll need to sell about $17 million worth of this crap - about 7,000 boxes?

Are they insane or simply detached from reality?

Cointerra has its own private farm packed full of their previously manufactured (garbage) HW.  I would think they don't need to money for NRE, but why spend your own when there are more than enough idiots out there?

Maybe someone should give them Spondoolies web address and suggest they buy some SP35's? Makes much more sense, dontcha think?
legendary
Activity: 1904
Merit: 1007
September 21, 2014, 04:49:22 PM
#71
And where did they get (or plan to get) the $7 million of NRE from?
They are hoping that the bitcoin community will give them a free loan... aka pre-orders.

So assuming a margin of 40% they'll need to sell about $17 million worth of this crap - about 7,000 boxes?

Are they insane or simply detached from reality?

Cointerra has its own private farm packed full of their previously manufactured (garbage) HW.  I would think they don't need to money for NRE, but why spend your own when there are more than enough idiots out there?

Exactly. Long live Inaba's scamming spirit!
legendary
Activity: 1512
Merit: 1000
September 21, 2014, 04:48:20 PM
#70
And where did they get (or plan to get) the $7 million of NRE from?
They are hoping that the bitcoin community will give them a free loan... aka pre-orders.

So assuming a margin of 40% they'll need to sell about $17 million worth of this crap - about 7,000 boxes?

Are they insane or simply detached from reality?

Cointerra has its own private farm packed full of their previously manufactured (garbage) HW.  I would think they don't need to money for NRE, but why spend your own when there are more than enough idiots out there?
sr. member
Activity: 441
Merit: 250
September 21, 2014, 04:43:29 PM
#69
And where did they get (or plan to get) the $7 million of NRE from?
They are hoping that the bitcoin community will give them a free loan... aka pre-orders.

So assuming a margin of 40% they'll need to sell about $17 million worth of this crap - about 7,000 boxes?

Are they insane or simply detached from reality?
hero member
Activity: 700
Merit: 504
Run a Bitcoin node.
September 21, 2014, 03:46:53 PM
#68
And where did they get (or plan to get) the $7 million of NRE from?
They are hoping that the bitcoin community will give them a free loan... aka pre-orders.
sr. member
Activity: 441
Merit: 250
September 21, 2014, 02:49:02 PM
#67
Leaving aside the wholes issue of preorders, a machine design using 16nm chips running at 0.3J/Gh is a piece of shit, period. And don't forget that it 'can be powered with an internal or external power supply rated at least at 1350 watts of continuous power', so it's not 0.3J/Gh at the wall, by the time you include the PSU efficiency at full load it's more like 0.37J/Gh - over 20% higher.

It's just another bog standard cell based design which will be a nightmare to product engineer into a reliable system. No vision, no innovation.

And where did they get (or plan to get) the $7 million of NRE from?
donator
Activity: 714
Merit: 510
Preaching the gospel of Satoshi
September 20, 2014, 08:50:06 PM
#66
Considering the amount of money they are spending on Google Ads, I guess they objective is to trap enough newbies into their presale scheme.
Their behavior seems to be modelled after BFL, and that is very worrying.
In fact, BFL and CoinTerra are the only two companies  (afaik) that have class action lawsuits as a common denominator... And with outlandish claims that they are powering 15% of the bitcoins network.
Lots of money seems to be flowing to advertising and on cranking the rumor mill.

BBB rating of CoinTerra: F
http://www.bbb.org/central-texas/business-reviews/virtual-currency/cointerra-inc-in-austin-tx-1000104440/
sr. member
Activity: 339
Merit: 250
September 20, 2014, 05:02:04 PM
#65
This is a highly speculative machine. We will see.
Pages:
Jump to: