Quote Originally Posted by JumpingJack View Post
Actually, yes it does.... +1 bin for 3 or 4 cores engaged, +2 for 1-2 cores engaged (with respect to Intel cores). The Lynnfield offer even more in the single and dual threaded scenarios. AMD's first implementation will run nominal with all cores engaged, and bump with 1-3 cores engaged. Either solution is absolutely a step in the right direction.

The move to multicore is simply a different way to take advantage of Moore's law, the physics, however, limits the clock speed not by Fmax of the design/process, but the practical limits of power and cooling. To fit a quad core, for example, within a reasonable thermal envelop, the processor is volted and clocked lower than what would otherwise be capable.

This is fine, since performance is extracted via thread level parallelism, in fact efficiency can dramatically increase if the task can be well threaded. However, Amdahl's law kicks in or some algorithms just cannot be made parallel at the thread level, as such these situations suffer with respect to the potential because any one core is clocked for the thermals.

Essentially, which would you rather have for a single/dual threaded app, a 3.4 GHz dual core or a 3.0 GHz quad core? Naturally, the best situation would be the higher clock.

So designers face a challenge, how to bring more and more cores to the masses but still yield performance on legacy single threaded applications or provide support until software can catch up.... the very natural thought is, if all but one or two cores are idle, why not just kick up the clock speed of the active cores... hence turbo modes.

It is a great idea, works as it should and gives you the best of both worlds, great performance in highly threaded environments, and a boost in performance over what could be had in lightly threaded environments.
If only the windows task scheduler was smart enough to use that potential...