more on this from a "Cell" perspective

quote from the article:
"For example, the nVIDIA GeForce 6800 Ultra, recently released, has been observed to reach 40 GFlops in fragment processing. In comparison, the theoretical peak performance of the Intel 3GHz Pentium4 using SSE instructions is only 6GFlops." [GPU][
http://www.scienceforums.com/showthread.php?t=2101

It is true. GPU development has gone down the path of RISC and I have come across an article where a database search engine was recompiled to run on a GPU and the performance was remarkable.

There are other instances of software being recompiled to run on the GPU.

I wonder how far this can actually be taken...esp now with the dual-head cards in parallel ?

Can you get a complete OS to run on a GPU and still talk to all the devices ? Will the mobo become just a BUS ?