5 Things You Didn't Know About GPGPU

Posted at: Tuesday 7th August 2007 by James Morris

So you’ve probably heard about GPGPU and have a vague idea about what it does. But where did it come from and how does it really work? Here are five things you might not have realised about GPGPU technology.

It may read like an acronym created by someone who’s spent too long listening to ‘My Generation’, but GPGPU is a technology which will be increasingly important over the coming years.

The concept isn’t hard to grasp. A modern CPU is a very clever piece of silicon, but a modern GPU has been honed to do its specialised tasks an order of magnitude more quickly. Those tasks originally revolved around rendering 3D graphics as fast as possible with the best quality available. But behind the pretty visuals lie complicated geometric calculations, so the same silicon and programming concepts can also be harnessed to execute all manner of alternative functions. This is General Purpose computation on GPUs, or GPGPU for short.

Essentially, anything which can be broken down into repetitive algorithms could be accelerated by a graphics card. The likely candidates include physics calculation, scientific modelling, financial forecasting, and other intensive data-processing tasks. ATi first brought GPGPU to public attention with its Stream Computing initiative, and we reported on how ATi’s graphics cards could be harnessed to accelerate folding, a scientific modelling application analysing how proteins are assembled. However, GPGPU is not being investigated solely by ATi. Nvidia GPUs can do it too, and Intel is also believed to be exploring the territory with its Larrabee initiative. 


So yes, you’ve probably heard about GPGPU and have a vague idea about what it does. But where did it come from and how does it really work? Scroll down and check out five things you might not have realised about GPGPU technology.