1.
Mr. Brookwood ...
"On any wafer, there will be many instances where two, four or even eight contiguous processors are all defect free. Intel needs only to keep these pairs of known good die together, dividing the wafer into 250 dyadic chips" etc.
This idea of Mr. Brookwood on how Intel could have demoed a non existing single chip dual core CPU is a bit dumb and reveals a large helping of ignorance about what CPUs are and how they are designed and manufactured.
As a way to get 2 or even 4 cores on the same piece of silicom of course it would work, but the resulting piece of silicom would be totally useless because it could not function.
Why? Because there would be no way to interconnect the cores to each other, to the arbitration logic (non existing) and to the external bus.
The only way this could work would be for the cores to be designed as complementary adjacent pairs, allready interconnected to each other, to an included arbitration logic and with a global external interface compatible with some CPU socket.
But then we would be in the presence of a true dual core solution. Wouldnt we?
Regards
Santos Costa
"On any wafer, there will be many instances where two, four or even eight contiguous processors are all defect free. Intel needs only to keep these pairs of known good die together, dividing the wafer into 250 dyadic chips" etc.
This idea of Mr. Brookwood on how Intel could have demoed a non existing single chip dual core CPU is a bit dumb and reveals a large helping of ignorance about what CPUs are and how they are designed and manufactured.
As a way to get 2 or even 4 cores on the same piece of silicom of course it would work, but the resulting piece of silicom would be totally useless because it could not function.
Why? Because there would be no way to interconnect the cores to each other, to the arbitration logic (non existing) and to the external bus.
The only way this could work would be for the cores to be designed as complementary adjacent pairs, allready interconnected to each other, to an included arbitration logic and with a global external interface compatible with some CPU socket.
But then we would be in the presence of a true dual core solution. Wouldnt we?
Regards
Santos Costa
[Posted by: S.Costa | Date: 09/23/04 02:34:38 PM]





