As a research project, I will be comparing the performance of certain types of calculations done on a gerneral procesing unit as compared to a graphics processnig unit. To acces the graphics hardware, I will be using OpenGL and writing a shading language to make use of the GPU's fragment processors. I'll post my progress here, but in the mean time, you can view my proposal.
* originally I was going to compare the CPU to a physics processing unit called PhysX, but due to the unavailibility of the SDK, my project has changed.