Tuesday, March 21, 2006
Tejas failed because of Terahertz transistor
Tejas, the biggest failure in Intel's history that happened last year, was the most mysterious
failure ever. No details, no explanation, only a kneeling of Intel's CEO and an excuse : We are in the transition towards a new architecture. In that moment that should be enough apologise, but actually, Intel's "new" architecture is not a new architetcure at all, but old one optimized for low power consumption.
Lets go back why Tejas Failed? Intel had plan to transit to Low-K based Terahertz transistor
in 2005 and demoed that achievement at IDF in 2001. However, Low-K had many integration problems and though it reduce leakage by more than 100 times it introduce its own ones. I think other things might be solved, simulations for Tejas have been done and everything was ready except the most important thing: reliably functioning Terahertz transistor.
Kneeling New Architecture was only Intel's backup move after Tejas failure. But is it Intel really
ever able to produce functioning Terahertz transistor? Yes, but not before 2009/10. Because
it missed its latest term target to do it in 45nm technology. We could see Terahertz transistor
even in 2008, but only in SRAM test designs . Yes, SRAM leakage can be reduced by switching off unused blocks of cache memory, like it was done with Yonah, but hardly with processor and still keep its performance.
Leakage current can rise by a factor of two to three for each process generation. "The next generation 0.10 and 0.065 micron looks horrible from a leakage point of view. We almost have to employ sleep transistors at 0.065," said Desmond Kirkpatrick, a CAD researcher with Intel.
Lets make a brief chronology: Geek.com first made an red alert on Feb 10 2003
10Ghz CPU aren's that far away.
However, instead of 10Ghz CPU, after that we have seen only Intel's CEO Mea Culpa on October 9th 2004. And finally, New Kneeling Architecture revealed at the latest IDF: Merom, Conroe, Woodrest.
The whole 2 years after Barrett's famous purifying kneeling gesture in 2004.
And what has happened in the meantime? Only one day later Dell corporation screamed:
Dell CEO says people don't want AMD chips
But, as destiny frequently turns events in reversed, instead of above statement,
WILL WE SEE SOON DELL'S CEO IN THE SAME KNEELING POSITION?
Let me instead of any conclusion. one famous historical citation:
Fascism should more properly be called corporatism, since it is the merger of state and corporate power - Benito Mussolini
Remains to be answered which one state exactly in this case?
After kneeling, we have seen Intel's new architecture development
departure to Intel's design center in Haifa, Israel and gargantuan $4,4B investment in a
new Intel's 45nm fab in Israel ($500 on average per each citizen of Israel).
No chance for more SCREW UPS at Intel?
BUT ,I AM VIVIDLY INTERESTED WHY PEOPLE DON'T WANT GOOD CHIPS?.
And vice versa?
That is the task BADHARDWARE blog should resolve in the future.
Because so much people actually don't want good chips, quite justified
I expect a lot of my blog fans.
This post link
failure ever. No details, no explanation, only a kneeling of Intel's CEO and an excuse : We are in the transition towards a new architecture. In that moment that should be enough apologise, but actually, Intel's "new" architecture is not a new architetcure at all, but old one optimized for low power consumption.
Lets go back why Tejas Failed? Intel had plan to transit to Low-K based Terahertz transistor
in 2005 and demoed that achievement at IDF in 2001. However, Low-K had many integration problems and though it reduce leakage by more than 100 times it introduce its own ones. I think other things might be solved, simulations for Tejas have been done and everything was ready except the most important thing: reliably functioning Terahertz transistor.
Kneeling New Architecture was only Intel's backup move after Tejas failure. But is it Intel really
ever able to produce functioning Terahertz transistor? Yes, but not before 2009/10. Because
it missed its latest term target to do it in 45nm technology. We could see Terahertz transistor
even in 2008, but only in SRAM test designs . Yes, SRAM leakage can be reduced by switching off unused blocks of cache memory, like it was done with Yonah, but hardly with processor and still keep its performance.
Leakage current can rise by a factor of two to three for each process generation. "The next generation 0.10 and 0.065 micron looks horrible from a leakage point of view. We almost have to employ sleep transistors at 0.065," said Desmond Kirkpatrick, a CAD researcher with Intel.
Lets make a brief chronology: Geek.com first made an red alert on Feb 10 2003
10Ghz CPU aren's that far away.
However, instead of 10Ghz CPU, after that we have seen only Intel's CEO Mea Culpa on October 9th 2004. And finally, New Kneeling Architecture revealed at the latest IDF: Merom, Conroe, Woodrest.
The whole 2 years after Barrett's famous purifying kneeling gesture in 2004.
And what has happened in the meantime? Only one day later Dell corporation screamed:
Dell CEO says people don't want AMD chips
But, as destiny frequently turns events in reversed, instead of above statement,
WILL WE SEE SOON DELL'S CEO IN THE SAME KNEELING POSITION?
Let me instead of any conclusion. one famous historical citation:
Fascism should more properly be called corporatism, since it is the merger of state and corporate power - Benito Mussolini
Remains to be answered which one state exactly in this case?
After kneeling, we have seen Intel's new architecture development
departure to Intel's design center in Haifa, Israel and gargantuan $4,4B investment in a
new Intel's 45nm fab in Israel ($500 on average per each citizen of Israel).
No chance for more SCREW UPS at Intel?
BUT ,I AM VIVIDLY INTERESTED WHY PEOPLE DON'T WANT GOOD CHIPS?.
And vice versa?
That is the task BADHARDWARE blog should resolve in the future.
Because so much people actually don't want good chips, quite justified
I expect a lot of my blog fans.
This post link