As always zarck,

The use of GPU's, be it using OpenCL or CUDA is up to the scientists, and as far as I know, porting an application to CUDA isn't exactly easy and may yield the desired benefits.

If a project would come along with cuda only, think it would get in notwithstanding the resistance, and lots of mewing in the amd/ati camp no doubt of being left out. Of course an IBM project running cuda only could be interpreted as an endorsement and the nxt nvidia add highlighting this as "see they too agree we have the best solution for computational biochemistry"

I think it is more likely that WCG would not divide their crunching community like that. My opinion, of course.

I will support WCG whichever way they go, Cuda or OpenCL. Only Nvidia works on Cuda. ATI works better than Nvidia on OpenCL and we heard lots of mewing from the Nvidia guys about that during HCC1. Some will be unhappy whatever happens, if GPU crunching happens here again.

And where is this proof? I have yet to see evidence of this to make such a broad general statement of this sort. From what I have read, it is the opposite...

You are saying that AMD on OpenCL is better than nVidia on CUDA? I would agree AMD is better than nVidia on OpenCL. I would agree AMD is better than nVidia in general for dual precision when comparing over all options available to the general consumer. However, I too have been running GPU's since pretty early in the game. I have not seen where AMD on OpenCL outright performs better in all areas compared to nVidia running CUDA. There is a good chance that if you find a situation where this is true, it is probably the scientists poor coding skills. The exception again would be dual precision work because nVidia cards are more crippled for DP work in general. I would also caution making such statements because CUDA (according to everything I have read) is a lot easier to code for and is more adopted by academics. So, depending on how you are defining "significantly better" is a pretty broad grey area. If you mean better as in smarter choice to adopt because more platforms can use it, then I would also agree. But I do not agree that AMD on OpenCL is better than nVidia on CUDA when only comparing those two head to head.

Edit: also take note that most of the comparison websites are not comparing abilities for DC'ing but rather gaming which is quite different. You also don't point out that the Titan would only be the competitor if talking dual precision projects. the GTX 780Ti is a couple hundred bucks cheaper though still significantly more expensive than the 7990. You will also find that many of the DC'ing option we have right now are still using older CUDA on newer cards. So, again, please show the example you are using to prove the statement you made...