Originally Posted by Tex1954
I got a crap load of GPU tasks taking anywhere for seconds to a couple minutes... no long ones yet I can see... still have about 50 cued up...
That is cause the WU's run both ways CPU or GPU and he has his wu issuer giving out WU based upon availability if there are no long tasks for GPU it will give out progressively shorter and shorter ones. (meant for the CPU's)
Me, I was getting WU's ranging from a few seconds to 20+ minutes.
Not that my equipment couldn't handle it, it is clear his servers cannot. He refuses to unleash the project in a clear GPU to CPU split mode. (cause the GPU work will run out fairly quickly leaving it as a CPU project)
And clearly stated that he will not bundle the tiny WU's for GPU consumption despite many suggestions for such to make his WU allocation system easier for the GPU crunchers....
His final suggestion was to run another GPU project along side his which I interpreted to mean we are applying too many GPU resources to his project. as his project was going to finish in two months at the current rate of production.
I never got an answer on if there will be another project after the COD database is completed.
So I wished him well, complemented him on a nice project roll out, and moved my GPU's elsewhere. I didn't want to cause him too much angst in finishing his project well before he originally estimated.
I think many others got the same hint given the projects serious decrease in daily production.
It's a small project, may last a few months, and any serious application of GPU power will complete it way before the admin intended or has the hardware to take advantage of.
That's why I left it.