Tuesday, 22 July 2008

Harnessing "hidden" processing

So the big processor chip manufacturers will show us impressive roadmaps for more cycles for less consumed power into the future.  And this together with multiple cores will significantly enhance the experience users get from future devices.  But for those gadgets with high definition displays, there is another strategy which will also add to the processing power equation.

The graphics chip (GPU) in such devices is extremely powerful in terms of processing specific types of data, painting billions of pixels per second on to computer screens.  But there is spare capacity in this task and so providing the data fed to it is suitable and techniques for sharing the processing within the program code are adopted, the GPU could become the accelerator for these devices.  Indeed Apple has announced its "Open-CL" initiative aimed at supporting this type of programming through future versions of its OSX operating system.  Given that GPUs can be ten times more power efficient than general purpose CPU chips, the impact on device performance could be highly significant. 

No comments: