The main problem with a CUDA client is OpenGL! CUDA only runs on Nvidia cards NOTHING else, there are a TON of other video cards out there and by ONLY making something work on Nvidia cards without Nvidia support is not very smart or a wise use of your time and resources. OpenGL is the new video card open source standard that is coming. It is out but not holey supported by all yet. When it is fully implemented and supported by all then most Boinc projects will settle down and exchange info on the hows, whats, etc and we will be crunching with our video cards too. Right now alot of projects are just waiting for the dust to settle down a bit instead of putting resources into something they know will be useless in a couple of years. Folding@Home made it work, but they have two totally separate versions and neither is Boinc supported.

If I were WCG (thus IBM), I'd not want to be seen as endorsing a single video card product, so there's logic of looking at a wider applicability with OpenCL. That's not to say that the first offering might be an nVidia release with HCC, but if you digest the words of bbover3, if WCG publish a plan it will be one that will be clear on the broad forward and no opportunity for a single card maker to advertise "we are the chosen". If WCG goes life, then I'll look at products and drivers for chosen platform... something that does not scream 'Burn' and does 250 watts an hour... it's still volunteer computing using idle/spare cycles... something that does not land inadvertently on the desktops of a 1000+ partner and messes up things in the relationship.

Only 3 Years?
One of my computers is 6+ years, 24/7 operating environment (I use it since Grid.org) with no break on activity for hard drive or anything else there and it still crunching nicely. I guess I'm lucky. The secret (everybody is telling me) is to not close your computer at all, if possible. Once is rebooting it might have electrical issues.

Just today I was reading an article in Computer Power User (CPU) examining CUDA applications. It mentioned Folding@Home but concentrated on video programs. One application ran using only 13% of a CPU, but most used at least 60%. Conclusion - it will take a while before the commercial programmers learn how to make good use of CUDA. My take on this subject - until HCC or another project has a CUDA version of their project debugged and running well on a GPU, WCG will not be feeling any pressure. I expect that we will be able to move fast then. Until we reach that point, we can prepare in a slow and deliberate manner since speeding up will not accomplish anything without a project.

Lawrence

I get that WCG is not currently supporting a CUDA aware version of BOINC because they are not hosting any projects that currently use CUDA. The issue I see is that people are told they have the wrong version (sometimes even when that has nothing to do with the issue at hand) if they are running a 64 bit version, CUDA aware version, or anything other than exactly what WCG's vision of what the BOINC world should be. Not everyone who crunches at WCG, only crunches for WCG.

ps. You bet I would I like to see CUDA science apps at WCG, I have a GTX295 ready to crunch away.