Hi there! I have been on here for only a few days now but really want to operate at maximum efficiency. I am not home much so I basically have my computer crunching all day long. I have a 4770K at 4.3GHz and 2 very highly overclocked GTX 770 4GB cards (1330MHz on core and 4300MHz on memory). I am running 4 tasks per GPU with .25 CPU to go along with each so I have 2 CPUs dedicated to GPUs and the other 6 CPUs are for CPU processing. Memory usage doubled as bumped up tasks per GPU (duh) and it is only at 1850 MB usage per GPU and temperatures don't go above 41C and CPU is at 65C at most. Would it be to much to push even further or will I run into the position where it starts to bog down my speeds? Open CL tasks usually take 6-20 minutes right now with 8 GPU tasks going and the CPU is working on Cosmology@home. I just don't want something bad to happen. Also, is there a way to have BOINC switch projects from time to time? I have to manually suspend one to get the other running like if I want to work on SETI.. bummer but not a huge deal. Thanks and hope to become a regular on the forum!
Josh Frechem

I changed it to run 5 applications on each GPU at a time with .2 CPU per application and it slowed down the process by 2 minutes so those 2 extra applications cause 22 minutes extra over 10 applications and 2 applications with only 4 apps/gpu would be about 22 minutes so it appears that once you hit 2 WU per gpu and reach 100% load at all times WU/core seems to scale linearly so you don't lose efficiency but you don't really gain. You do use considerably more VRM though ( it is now using 2225 MB per GPU)

I changed it to run 5 applications on each GPU at a time with .2 CPU per application and it slowed down the process by 2 minutes so those 2 extra applications cause 22 minutes extra over 10 applications and 2 applications with only 4 apps/gpu would be about 22 minutes so it appears that once you hit 2 WU per gpu and reach 100% load at all times WU/core seems to scale linearly so you don't lose efficiency but you don't really gain. You do use considerably more VRM though ( it is now using 2225 MB per GPU)

Yeah you might want to back off just a bit to one less unit at a time, burning up your gpu's is not an ideal situation. And since that one extra unit is now putting your gpu at the 100% number and causing a slowdown, you are probably not running it very efficiently anymore. This is not a race nor a sprint, Boinc is a long term thing, don't burn up all your resources in the beginning and then be sitting on the sidelines later on.

As for your other question about switching projects that is what Boinc does on it's own and there is little you can do beyond what you are doing to make it switch back and forth when you want it too. One thing you might try though is to have a smaller cache, that way Boinc might switch back and forth more often. The other thing you could do is to put one gpu at one project and the other gpu at a different project.