On Tue, 15 Mar 2005 18:05:02 +0900, David Cournapeau wrote
  Hi there,
     I think I am not the only one here to have heard about so called
 '3d ray tracing cpu' (e.g:
 
http://graphics.stanford.edu/papers/rtongfx/rtongfx.pdf, and a
 recent news on /. about some demos at CeBit ). There also were some
 emails here about using GPU for sound synthesis recently, mainly
 about the shortcomings of those technics (memory bandwidth problems,
 difference of rate scales, if I understood corectly).
     Unfortunately, I know nothing about computer graphics, I don't
 know exactly what this ray tracing cpu technology is (maybe
 vaporware ?), but I wondered if this can be used for audio
 technology, e.g RT reverbs, etc... Does anyone have an idea about
 the possible usage of this kind of chips for audio processing ? 
GPU's can be used for general purposes, but I'm not sure how suitable they
would be for audio work - the reason for this is the readback latency for
getting data back from the card, they are currently highly optimised for
display only (for applications like games). This is changing though, with
nvidia working on PCI express, which should speed this up considerably.
This is also interesting: 
http://graphics.stanford.edu/projects/brookgpu/
cheers,
dave