Stand back way back in the proposers queue... you are person 5,001 and BOINC 6.6 already can use the GPU, except WCG has no project that uses this still not fully mature feature [by WCG standards].

I think the reason why the WCG does not presently (June 2009) use NVidia's CUDA or some equivalent ATI Graphics cards crunching software is down to project management and planning.

To utilise graphics cards for processing takes a lot of project planning. Many existing projects were planned some time ago, and some projects are in their second or third stages, they may have began years ago - before graphic card crunching really caught on. The number of CUDA capable devices is rising rapidly, but a couple of years ago it was really just getting going.

I’m sure that the WCG would like to support its projects with such technology, and is looking into it, but there would have to be a new project designed to utilise graphics card processing.

When it does come it will be a fantastic asset to the WCG, and I am sure the WCG will introduce it in a very professional manor, unlike some alternative projects that published CUDA projects before they were available, or release alpha projects and pseudo CUDA tasks to the general public.
I’m quite happy to wait for the WCG to do it correctly.

It doesn't matter how fast you drive, if you crash your car you don’t get there!

Ever heard of hardware diagnostics?
For every problem there is a solution.

If GPU-Grid can overcome the problems of CUDA so can the WCG.
My basic CUDA graphics card (Nvidia GeForce 8800GT with 112 shaders) can easily do twice the work of a high end desktop CPU (Phenom II 940, 4GB RAM, 1TB, 64bit platform). Last month my CPU produced 25000 WCG points, and the 8800GT did over 50000 for GPU-Grid. I cannot get a faster CPU for my system, but I can get a much faster GPU!

Mid range Graphics cards (GTX 250, 9800GTX, 9800GTX+) get about 30% to 50% more results than my card. High end cards (9800 GX2) get about twice the results of an 8800GT, and a top card (GTX 295 with 480 Shaders) can easily get three and a half times the results of my card (that’s 7 times the work my CPU can do on one graphics card).

I could have 2 in my system and many systems support 3 or 4 PCIe cards!

A gamers system with 4 GTX 925’s would produce about 20 times the results of the top i7 CPU.

The latest CUDA cards to hit the shelves are the S1070 and S1075. These have 960 shaders. I can only guess that their performance is about 10 times that of my 8800GT card.

At present GPU processing is pulling away from CPU crunching rapidly, so it’s worth the effort to bring CUDA to a WCG project. However, not until it is up and running correctly! No put-off false starts, disastrous alphas, Pseudo CUDA, hybrids or intermittent workloads please.

ehh, there are issues with the graphic card crunching non the less. I've heard that some of the main issues are the graphics drivers aren't good enough and they don't have process prioritization and the ability to limit use by a certain process. Hopefully it'll get better and WCG becomes more interested. Of course I probably couldn't contribute because my graphics card is awful XD