Let me create a simple example on what I mean by the title of this page. Imagine the two following systems:
|System A||System B|
|Idle Power Requirement||50 W||80 W|
|Max Power Requirement||120 W||120 W|
|Performance||State of the art||Average|
Let's now think of a typical workload, such as a video editing task. While you drag and drop video fragments in your video editing application, for let's say 30 minutes, both systems will mostly remain close to their idle power requirement, as only few of your actions might really stress the hardware. Once you finished the manual work, the systems have to encode or transcode the video into the target video format and container. Obviously, any system out there that doesn't have hardware to accelerate video processing will be quite busy working on the video.
This is where the maximum power requirement makes its appearance, and where you should carefully take system performance into consideration. If system A were considerably faster than system B, it would complete the video transcoding much quicker, so it will go back to its low-power idle state faster, which in the end allows for the greatest power savings.
In this example, system A has the potential to win the comparison, because its maximum power requirement under load equals the power requirement of system B, but system A's idle power is much lower. Clearly, returning to an idle state once the work is done will result in an excellent power consumption result. System B will take longer to process the workload, while also requiring more power while working on it. This increases the total power consumption over time, as measured in watt hours (Wh).
If you look at the entire video editing and creation process, system A would still consume less power, even if its maximum power requirements were higher. Whether or not using a faster component with a higher power requirement makes sense clearly depends on the workload. Applications that scale well can terminate earlier and put the system back into a more efficient energy state more quickly.
This is why I'm excited to see the next generation of quad core processors. And I'm not referring to Intel's 45 nm dual-die Core 2 Extreme - which will arrive before the end of the year - but AMD's Phenom X4 (aka Barcelona), and Intel's 2008 quad core that will be based on the Nehalem micro architecture. Both products are capable of modulating clock speed and switching off individual cores when they're not needed. These future processors will be perfectly suited for changing workload conditions, providing four-core performance for demanding workloads combined with close-to-single-core power requirements when there's nothing to work on.