I usually switch between Folding at Home and BOINC each month top participate in each.  Well I was in the hospital during the event so I could not get my computer to run this program while I was out.  When I got home I was too tired and felt to bad to try.  Well yesterday I thought I would try.  I did about a week of Folding at Home and now want to some GPUGrid, PrimeGrid and World Community Grid work.  Started up BOINC and GPUGrid seems to be working fine, but I cannot get any projects for Prime Grid or for World Community Grid no matter what.  Anyone have any idea why this would be happening?

I finally started to get some work, but I had to remove and then reattach the project.  This was the same thing with WCG.  But I only get a few projects, much less then I even did.  GPUGrid has been giving both CPU and GPU tasks.  Is that the reason?

Yes... basically GPUGrid was owed run time per BOINC Manager trying to give resource shares according to how you have them set. Since there was CPU work, it filled your CPU cores to help give it the time it was owed. When you reset the projects it in essence reset the clock for the other projects as if they were new.
 
Since GPUGrid doesn't have CPU work very often, your CPU projects were racking up a ton of time while your GPU projects were fighting to match that with only the GPU's you had. But, in this case, GPUGrid was finally fed with more than GPU. It isn't a priority problem rather than just letting it run and it will balance again. The CPU work for GPUGrid will dry up again too. Changing the resource share will only make other GPU projects get more share as well...

Keep in mind that the resources will still be tallied like before. So, unless you continue to run an equal amount of time on GPUGrid, it will happen again. In the future, just let the client do its thing. It will give them equal time based on the resource share you have assigned. It may look like you aren't doing anything at a project but that is because it has already used its share of time up. You can also use an app_config to limit the number of work units you get from a project at a time which is another way to guarantee project splits. However, with GPUGrid, they have a multi-threaded CPU app. So, unless you also limit the number of cores/threads it is allowed within a config file, it could still theoretically happen. Just some food for thought.