Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
![]() |
World Community Grid Forums
![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
No member browsing this thread |
Thread Status: Active Total posts in this thread: 7
|
![]() |
Author |
|
Jason Ellis
Cruncher Joined: Apr 3, 2007 Post Count: 4 Status: Offline |
This maybe one of those super secret hand shakes that the cc_config file controls..
I know that between my two graphics cards I have over 300 cuda cores.. can I divide them up and allocate say 50 cores to a job or some such? Or is the division right now only by card? |
||
|
Dataman
Ace Cruncher Joined: Nov 16, 2004 Post Count: 4865 Status: Offline Project Badges: ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
WCG does not support GPU processing. There is a lot of GPU information at BOINC or GPUGrid.
----------------------------------------can I divide them up and allocate say 50 cores to a job or some such? No. Or is the division right now only by card? Yes. ![]() ![]() |
||
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
WCG isn't the only projects I'm associated with for that very reason.. Milkyway, Einstein, Docking, SETI..etc.. they all will use CUDA..
BOINC itself determines the split and yeah appears to be on a per card basis. |
||
|
Dataman
Ace Cruncher Joined: Nov 16, 2004 Post Count: 4865 Status: Offline Project Badges: ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
First, credit were credit (or dog year points) is due. I totally ripped this off from a post by Kevin (skgiven) in another forum.
----------------------------------------Introducing the ultimate crunching machine. The new EVGA GTX9990 super-duper overclocked Super-Collider FERMI: ![]() for the introductory price of $11,999 USD. If you act now, it comes with a FREE 10,000W PSU (just pay additional shipping and handling of $1,000 USD). Act now while supplies last. ![]() ![]() ![]() [Edit 1 times, last edit by Dataman at Mar 18, 2011 6:39:16 AM] |
||
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Getting a figmetive picture in my head of Berkeley having put in the table that GPU XYZ has 300 cores and thus due ABC credit per second and a volunteer starts splitting the card over multiple tasks running concurrently... that would be very serious credit heaven. Till then, it's 1 blob in, 1 blob out.
LoL |
||
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
You can run more the one task on a GPU but not recommended since they will slow each other down hardcore.. GPU processing is alittle different since the task itself uses max power the application can support meaning one task may not use 100% GPU I seen tasks use anywhere between 50-80% GPU it depends on the size of the WU also I have noticed the throttle it too.. smaller WU will use the GPU less compared to a larger WU.. also trying to run more then one WU on a GPU is room for problems with both WU's..
|
||
|
sk..
Master Cruncher http://s17.rimg.info/ccb5d62bd3e856cc0d1df9b0ee2f7f6a.gif Joined: Mar 22, 2007 Post Count: 2324 Status: Offline Project Badges: ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
Multi-tasking efficiency depends on the GPU.
----------------------------------------GPU utilization depends on the GPU, the app and in some cases the operating system. I regularly see 98% utilization, but some projects only use about 47% - For Projects such as Einstein many users configure Boinc to run more than 1 task on their GPU. This can be done by building the app_info.xml and while it works for other projects (I've tested this) it does not significantly improve overall performance and often reduces it and causes GUI problems. It's thus not recommended by most projects. Some of ATI's forthcoming generation of GPU's are expected to support this more efficiently; to the extent that you could conceivably run two entirely different high utilization projects together with little or no loss. This would be an attractive and simple way to split your GPU research projects up evenly, but would at the very least require brutal surgery to Boinc. [Edit 1 times, last edit by skgiven at Jun 23, 2011 10:39:13 AM] |
||
|
|
![]() |