Intel bets on multicore before they go the way of the big three auto makers... Well that is old news, their battle with AMD turned out to be just a surge from an arch rival. It seems AMD will have to stick to the low end computing market for some time. Unless consumers decide to embrace cheap AMD will have a hard time competing in the Apple-Intel 'common sense computing' revolution.
intel has been on the muti-core path for the past half a decade with an aim to expand the performance of individual processors to 10s if not 100s of cores allowing for real parallel processing and programming. The real goal is to be able to provide uninterruptible computing by allowing each application to be processed on a separate thread. The below diagram shows Intel's hope that a "Thread-Aware Executive Environment" will allow the above 'high-level' applications transparent and efficient use of a massively-multi-core CPU.
The 80-core prototype is a Teraflop (one trillion floating point operations per second) research silicon which focuses on design specifications for low-power, core-to-core interconnects. This research is similar in scope to the IBM cell-processor presently implemented in the Sony PS3. Intel's research points out what many silicon manufacturers have already concluded, that while performance gains can still be made through frequency scaling there is a cost in terms of energy efficiency including cooling. In 1996 10,000 Pentium Pro processors consumed 500kW of power plus another 500kW of cooling while delivering a teraflop. In contrast the Teraflop Research Chip (TRC) delivered one teraflop of performance while dissipating 62W.
Intel expects that the use of more cores will change conventional programing to a Network On a Chip (NOaC) approach.
an 80-core chip made up of mostly floating point engines.
