GPUs are only capable of a subset of what a General Purpose Processor (that's a CPU) is, so coding for them requires reducing the problem at hand to a more restricted set of operations and memory limits than with a CPU based application. GPUs are also massively parallel by nature so they require a different set of programming skills and tools to get the most out of them compared to typical CPU programming. There are more "drop in" libraries becoming available to developers as time goes on that allow GPU functions to be inserted into regular C code without requiring special skills, but where it doesn't matter a great deal if a drill head is a meter or so off target because of a rounding error, biomedical work has no such tolerances for error. Medicine and poison are often only a question of dose.

Forgot the tongue in cheek smiley... the established, middle of the road cards with longer supported drivers. Same as in a business, the cows ensure continuity, and the stars [can] put the icing on, the dogs you'd like to loose and the question marks need lots more work before they contribute, in some uncertain future.
--//--

So? If I understand Kevin's post, I have to run HCC non-Beta CPU wu's in order to get GPU beta wu's?

Piccolo Passi.

--//--

I followed the procedure in Kevin's post and for my sins I received ~150 HCC CPU wu's that I will be aborting in the next few minutes and, of course, no GPU work. That is only for 7 cards on 4 machines. If I start the other 7 cards, I would lierally get ~300 wu's (cashe = 0.01 days) that I don't want for 80 CPU cores. If this is the case, I cannot participate in this Beta.