There’san articleup on CNET about Intel’s research into multi-core chips for things like on-chip SMP. One interesting thing they want to do with multi-core chips is use them to save power and reduce heat.
In “core hopping,” for example, a stream of calculations will jump from one microprocessor core to another. Localized heat generated by transistors during intense number-crunching can create “hot spots” that place a ceiling on performance, said Wilf Pinfold, technical director of microprocessor research at Intel labs. By rotating application processing, key transistors will stay cooler, heat will be dispersed over a greater geographic area and overall performance will climb.
I hadn’t actually heard of this technique before, and I’ll be interested to see how they combine it with SMT (simultaneous multi-threading). SMT is designed to keep more transistors busy at any given moment by executing in parallel on a single core code from multiple, concurrently running threads, instead of executing one thread at a time and switching between them really quickly in order to give the illusion of parallel execution. And now this “core hopping” technique seems intended to offer flexibility in the other direction: you can power down a functional unit and let it cool by shifting the work to another, identical FU on a different core, which means that there are fewer transistors working within the same core at any given moment. I think that both of these abilities–the ability to do more at once with a single core and the ability to do less at once with a single core–will make for a lot of flexibility in dynamically scheduling code and execution resources for the optimal combination of power consumption and performance appropriate to the task the processor is currently executing.
