How do you do. Suppose you have a Pentium III or more, plus a 3D video card. This video card can handle polygons, triangles... actually a lot of calculations, like the CPU does. Now is it possible to make this 3D processor work, through an API call, make some CPU job, and retrieve computed 3D work a bit later : not to the monitor, but to CPU back or into physical memory ? See what I mean ? A bit as what is/was done with FPU... As a summary, I make intense calculations with MMX that does not fit well with FPU, and I wonder if a 3D processor could be used in parallel. I am a newbie about OpenGL, as you see. Thanks

to some degree you can but it is a useless excerise as the bus transfer rate would any such usage pointless. Moreover the whole design of the cards is not made for that kind of usage. Learn how to use pipelining more efficiently instead.