Originally Posted by Angrybutcher
I just did a bit of searching as I had no idea what that was. I get the gist of it now, but unsure how to implement it or even if DistrRTgen supports it?
Well it should support it. Since it's a system environment variable of windows and the task runs in the environment.
If a gpu task requires X amount of cpu -> dedicating one thread on your hardware will make the gpu task faster, in theory. atleast it does on gpugrid.
Since it is the cpu that feeds the gpu with work. Let's put it this way, 2600K@5Ghz+GTX580@900core 4way SLI is faster than say E8400@4Ghz+GTX580@900core 4way SLI because 2600K can feed the work to the gpu's faster than the E8400 can.
You enable it this way:
E: Also you need to set in boinc = "On Multi Processors Use XX%" option to 50%.
After you have done that, you can only run two cpu tasks on your processor because you have dedicated 2 out of 4 threads in your CPU to your GPU's.
But that shouldn't matter since PPD difference on CPU vs GPU is so freaking huge.
If you do the swan sync and leave the setting at 100% then your GPU and CPU tasks will all suffer as there won't be enough CPU time available.
E2. Oh hey hey hey I came up with something.
Angrybutcher, look at http://boinc.freerainbowtables.com/ Your account and tasks, look at the cpu times. How much for you?
For me they vary between ~5.3 and ~9.6 seconds with avg being 7.8 seconds
And I have set swan_sync enabled.
Edited by lagittaja - 2/8/12 at 5:51pm