AMD's Fusion concept is completely different from Montalvo's concept. Fusion integrates what would normally be external devices into the CPU package, and those other cores still are only processing specific code designed for those cores. To start off with, graphics code, but eventually maybe game physics acceleration, or cryptographic stuff, or other things that I don't know of that business or research currently uses accelerator cards for. Montalvo is processing generic x86 instructions on every core, and directing the instructions based on some method of determining what needs better performance.

I don't see how they could possibly determine what processes need better performance though. I don't even understand how SpeedStep or Cool 'n' Quiet determine when they need to run at full speed, other than "the core is at 95% usage for more than 5 seconds so we must need higher performance". Montalvo's idea would seem to go beyond such a simple measurement, since it would need to judge how the code will perform on each type of core, and still can't be all that accurate since CPU usage may not translate to how the user perceives performance.

Also which core to use would depend on how long the thread is going to need the extra performance; turning on the better core to run the code which will only take 30ms could be more of a waste of power than just letting it run for 2 more seconds on the slower core. No scheduler can know how long a thread is going to need to run.

Do they have any chance of going fabless?

hmm...Montalvo is starting to make news already? a probable (more worthy) third party (contender?) in the industry?

this i wanna see XD