I don't know if this is a valid criticism, but it has occurred to me that any scientist who are capable of writing CUDA or other GPU based apps may not contact WCG simply because they know it isn't supported. I mean, unless there is some sort of survey that would tell you how many labs are interested in doing GPU based crunching, it could easily be a self-fulling prophecy.

I'm not questioning anyone's dedication, only the reasoning for not making it more of a priority. As I understand the argument, the main reason is that there are no projects that have approached WCG with a desire to use GPU based software. But if I am a scientist with a project that could be run on GPU's and that is my preference, I wouldn't even approach WCG since I would know from doing some cursory research that it is not supported.

If I had absolutely no other options, I might at that point see if I could get GPU software supported, but if the turn around would be unacceptably long, I would have to think of something else. While scientific progress may be slow, people aren't going to wait for many months if not years to get time on WCG. We're good, but we're not THAT good.

I'll be happy to see that. The 2.4GHz Phenom I'm on now does about 2.6GFlops. A new HD5870 does 2.6TFlops - 3 orders of magnitude faster. Granted it isn't as versatile as a CPU, but d*mn, 1000 times more flops . . . that deserves some respect.

Lot of good posts in this thread.
To the CA's;
I think the passion you see here is like when a lot of people wanted to see 64 bit computing being able to be incorporated into the WCG app.
Whether 64 bit or GPU computing what your seeing is desire to cut the time frames on doing the work.
I know it may sound confrontational at times but it's really not. Just passion my friends.
Now as much as I'd like to see both 64 bit and GPU computing added to WCG I can honestly say I tried GPU computing with GPU grid on 2 machines here.
What I got made no sense to me.
I put a cheap EVGA 9600GSO 384MB card in a 940 AMD system and it worked fine.
I put a EVGA 260GTX card in the Gainestown rig and it errored more than you could imagine until I finally gave up on it and that machine has nothing but top end parts in it.
Very strange so based on what I saw and also have read at the GPU grid forum I'd have to agree that it's not quite ready for "prime time"
I'm also concerned that when/if GPU computing gets perfected that we'll see lots of issues here.
Video cards maxxed out make LOT's of heat and not as many are familiar with this as are with dealing with overclocked cpu's.
You can't just say :' Stick a ThermalRight Ultra Xtreme HS on it and call me in the morning."..
Little more involved with the GPU's and then add in extra case cooling to whisk away that heat and what I'm seeing is many issues with Mr./Mrs. Jones with their new Dell XPS but not much knowledge of what's in there having their vid cards burn up.
VERY hard to kill a cpu but you can smoke a video card easily.
Something to think on anyway.

Anything to speed up the work is great. Increasing it by orders of magnitude is even better. With that potentialy awaiting it is just a matter of time! Need a X prize for it.