I have a working prototype of Bitcoin generation with OpenCL.  OpenCL is similar to OpenGL but it's for doing computation, not graphics.  Other similar technologies are CUDA (NVIDIA) and DirectCompute (Microsoft).  OpenCL is Apple's version of it but it is available for Linux and Windows as well.

I posted a Mac OS X binary package in my other thread.  I will create a patch and write up instructions for each platform this week as I have time but I wanted to find out who if anyone is interested in this even.

Please reply if you're interested and what OS/Video card you have.

Supported video cards:
NVIDIA: http://www.nvidia.com/object/cuda_gpus.html
AMD (ATI): http://developer.amd.com/gpu/ATIStreamSDK/Pages/default.aspx#two

Just to give people an idea of the benefit..
In my development computer I have an Intel E8600 CPU which runs at 3.3ghz normally.. I have it clocked at 4.1ghz
With 2 threads running the normal way I get about 1800k iterations per second in the bitcoin miner.
With an NVIDIA 8800 GTS (G80 GPU) video card I get around 3300-3800k iterations per second (varies with tuning of the code).

I find the best total performance comes from running one CPU thread and one OpenCL thread for the GPU (which does a bunch of work in parallel).  I get about 5000k/sec total that way on my development machine.  If I use 2 CPU threads it slows down the overall results to about 3600k/sec because the CPU is tied up computing and not pumping data to the GPU fast enough.