There are a number of us with either slower GPUs or busier computers that cannot complete WUs in the 5 day window alloted to them. Here is what I had to resort to, I hope that it helps you set up your strategy if you're in the same boat.

In may case, the computer that I am running GPUGrid on is busy 5 days a week for up to 12 hours. In order to ensure that my WUs don't age out, I do the following.

On Wed morning, I allow the GPUGrid task to get new tasks (Projects tab, Allow new tasks button) and update the project. I get two WUs as a result. I immediately abort one because both will not complete. Then I set the GPUGrid task to No New Tasks so it won't pick up a second one.

The sole WU will now easily complete within the 5 day window because it will get two full days on the weekend. During the Monday and Tuesday slack hours, the GPU will be idle - starting a new job at the beginning of the week may not allow the GPU enough free time to complete on a M-S run. On Wednesday, I repeat the procedure.

Why do I bother with this? Because the 8 cores of my machine are processing other protein projects and the machine is running. It seems like a waste not to put the GPU to work also.

What little tweak from GPUGrid would help? Expand the window to 7 days - we would always see Saturday and Sunday in the processing of these long WUs. Of course, GPUGrid only downloading as many WUs as GPUs you're devoting to the project at a time would mean no manual intervention on my/our part.

Hope that reduces some of the shared frustration in not being able to participate in GPUGrid.

The sole WU will now easily complete within the 5 day window because it will get two full days on the weekend.

OK, it's a total waste of time.

The question is then, why in the world would you give credit when the results are discarded? Just how much credit is being awarded for results that you discard? Why don't you notify those processing discarded WUs that they are discarded?

Wouldn't the project be more transparent and honest if you simply didn't award the credit? You have the benchmark results for each system. Why in the world don't you put a lower bounds on the metrics and deny WUs to those systems?

It really appears that the whole GPUGrid project has issues. It's apparent that the group does not have experience in product management and requirements management. You are also not working well with the community based on the repeated threads on the same issues.

You probably don't understand these issues right now. When you guys get out of school and deal with real world projects, you'll look back and see how poorly you are managing the external aspects of your project.

Good luck with the project and results. As a former practicing crystallographer, I appreciate the work in protein structures and believe that, among all the BOIN projects, these are the most rewarding.

<disconnect>