There's been a lot of talk this year about the potential of parallel computing in the home. Nvidia, Intel and AMD are all pushing the idea of using the GPU on a graphics card to offload data-intensive tasks from the CPU.
The Nvidia GeForce 9600 GT, for example, features 64 programmable shader cores. For the most part, a card like this is designed for rendering high-level visual effects in video games like Far Cry 2 and Fallout 3. But the GPU's processing cores can also be reallocated for other tasks – i.e. calculating real-time physics (PhysX), decoding H.264 video or enhancing the quality of photos.
It's early days for multi-core processing. The majority of today's applications are built with single-core processors in mind. But the idea of using the processing muscle of a graphics card GPU has definite benefits for digital home applications.
The secret life of graphics cards
GPUs are ideal for math-heavy tasks – scientific systems featuring clusters of GPU chips are already outperforming CPU-only hardware in fields as diverse as molecular dynamics, ray tracing, medical imaging and sequence matching.






