I think that most of the GPU projects are short of work. Certainly POEM is, and even GPUGrid barely manages to keep up, though I have had no problems recently. And there is no point in piling on if a project is short of work, since any work unit you get is simply taking it away from someone else. It might actually decrease the overall efficiency of the project by placing an unnecessary load on their servers.

The only exception I know of is Folding@home, which manages to generate tons of work, maybe because of the nature of the problem and lots of researchers, but also because of their infrastructure. With their new Core_17 coming out before long, the AMD cards will get a big boost, especially the higher-end 7000 series. So I think you had better be prepared to shift around and not expect any one project to keep you busy, and to buy new cards with an eye on what is required in a given area.

But note that one potential problem with Folding is that it usually does not like to deal with dissimilar cards if you run two or more, at least with the current cores (though Core_17 might be better). However at the moment I am running HCC on an AMD HD 7770, and also Folding on a GTX 560 Ti on the same PC, so it can occasionally work; it is hard to predict. BOINC does not seem to have that problem, and I have had no particular problems getting different cards to work on various BOINC projects.

Based on current estimates, there is now approximately 25 days of new work remaining for the Help Conquer Cancer project. Once all new work has been sent out for the project, only resend work units will be available (work units not previously computed correctly or on time).

Seippel

My curiosity was more along the lines of WCG having a requirement of new projects having enough work to do to last a certain number of months. If a new project is willing to do it on GPU but finish in a month, then obviously WCG would turn it down in this scenario. However, the scientists could just leave it on CPU to get WCG to pick it up and then later introduce their GPU app thus finishing early. I was wondering if this logic is applied when promoting new projects. Obviously this is an extreme example, but there are many jaded by the sudden ending of projects recently. I for one am glad they finished early, so please don't think I'm being negative.