How do I make sure only "like" work units execute on the GPU?

Message boards : Number crunching : How do I make sure only "like" work units execute on the GPU?
Message board moderation

To post messages, you must log in.

AuthorMessage
Chris Adamek
Volunteer tester

Send message
Joined: 15 May 99
Posts: 251
Credit: 434,772,072
RAC: 236
United States
Message 1496444 - Posted: 28 Mar 2014, 4:52:22 UTC

How do I make sure only two APs or 2 two MBs execute simultanesously. Does it even matter if I end up with a mix running on a single GPU? I suppose that should have been my first question lol...

Thanks,

Chris
ID: 1496444 · Report as offensive
Cosmic_Ocean
Avatar

Send message
Joined: 23 Dec 00
Posts: 3027
Credit: 13,516,867
RAC: 13
United States
Message 1496447 - Posted: 28 Mar 2014, 4:59:00 UTC
Last modified: 28 Mar 2014, 4:59:37 UTC

Interestingly, I think I've read over the years that mixing the type of work being done on one resource tends to yield faster execution. The only real issue with GPUs is the amount of memory consumed by the tasks, especially when you allow multiple tasks to run simultaneously on one GPU.

A lot of people run 3 MBs but only 2 APs, and it gets kind of sketchy when they start mixing if you have the configuration set to pretty much the absolute maximum capacity instead of somewhat conservative to allow for some headroom.

I am by no means a GPU expert though, so the more seasoned veterans should chime in before long with a better/right answer.
Linux laptop:
record uptime: 1511d 20h 19m (ended due to the power brick giving-up)
ID: 1496447 · Report as offensive
Profile petri33
Volunteer tester

Send message
Joined: 6 Jun 02
Posts: 1668
Credit: 623,086,772
RAC: 156
Finland
Message 1496502 - Posted: 28 Mar 2014, 8:15:52 UTC
Last modified: 28 Mar 2014, 8:17:03 UTC

My machine runs 4 AP or 3 MB best. When mixed 2 AP 1 MB the MB runs slow. When 1 AP 2 MB all is okay. My AP is 0.25 and MB is 0.33.

I don't know how to limit to one type only.
To overcome Heisenbergs:
"You can't always get what you want / but if you try sometimes you just might find / you get what you need." -- Rolling Stones
ID: 1496502 · Report as offensive
juan BFP Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 16 Mar 07
Posts: 9786
Credit: 572,710,851
RAC: 3,799
Panama
Message 1496547 - Posted: 28 Mar 2014, 10:45:58 UTC
Last modified: 28 Mar 2014, 10:46:17 UTC

ID: 1496547 · Report as offensive
Chris Adamek
Volunteer tester

Send message
Joined: 15 May 99
Posts: 251
Credit: 434,772,072
RAC: 236
United States
Message 1496633 - Posted: 28 Mar 2014, 14:31:32 UTC - in response to Message 1496547.  

Thanks! Now a second question. Is there anyway to have different settings for for different GPUs in a single system. For example, in one machine I have 560 Ti and a 750 Ti. I'd like to run 2 AP's on the 750 a 1 on the 560. The same would apply to different command line settings for AP or even running CUDA 5.0 on the 750 ti vs CUDA 4.2 on the 560 ti?

Thanks,

Chris
ID: 1496633 · Report as offensive
juan BFP Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 16 Mar 07
Posts: 9786
Credit: 572,710,851
RAC: 3,799
Panama
Message 1496636 - Posted: 28 Mar 2014, 14:46:32 UTC
Last modified: 28 Mar 2014, 14:49:18 UTC

The only safe way i know to do that is by running 2 diferent instances of Boinc on the same host, since your 2 GPUs are from the same arquitecture (NVidias), something not easy to do and few realy knows how to do.

If you wish to try, there is a guide on Vypers site, but be aware, try only if you realy knows wath you are doing, it´s for (very) advanced users.

Here is the link: http://vyper.kafit.se/wp/index.php/2011/02/04/running-different-nvidia-architectures-most-optimal-at-setihome/
ID: 1496636 · Report as offensive

Message boards : Number crunching : How do I make sure only "like" work units execute on the GPU?


 
©2024 University of California
 
SETI@home and Astropulse are funded by grants from the National Science Foundation, NASA, and donations from SETI@home volunteers. AstroPulse is funded in part by the NSF through grant AST-0307956.