v0.42 with mixed GPUs

Message boards : Number crunching : v0.42 with mixed GPUs
Message board moderation

To post messages, you must log in.

AuthorMessage
Profile Zalster Special Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 27 May 99
Posts: 5517
Credit: 528,817,460
RAC: 242
United States
Message 1558687 - Posted: 18 Aug 2014, 12:39:39 UTC
Last modified: 18 Aug 2014, 13:04:58 UTC

I didn't post this in the Installer release notes as I wasn't sure it was the right place to ask this question so I make this thread instead. I'm running the v0.42 on all my machines but only 1 of them seems to be having this problem and I think it is because of the mixed versions of GPUs in there. I have 2 780s along side 2 750s This is the setting I use

-use_sleep -unroll 12 -ffa_block 12288 -ffa_block_fetch 6144 -tune 1 64 4 1

Window 7,
Driver 340.52

I'm running 3 APs at a time all all the GPUs. For the last 3 days, I've come back to find the system frozen and I have to do a hard reboot. It wasn't doing this before I installed the new optimized apps, so I am wondering if it's due to the settings and mixed GPUs that it is freezing? I know there was talk about the drivers in other threads, that might be something else I may have to try as well. Any ideas?

Edit..

Temp on 1 GPU is 71C with 100% fan, the other run 57C, 65C, 46C
ID: 1558687 · 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 1558692 - Posted: 18 Aug 2014, 13:10:08 UTC - in response to Message 1558687.  
Last modified: 18 Aug 2014, 13:12:20 UTC

First of all i´m not sure if they allready fix the 340.52 bug when you runs with -use_sleep check in the specific thread about that. Because that i still ussing the 337.88 driver.

Back to topic.

You have 2 options, configurate for the slower one (the 750) and use a less agresive parameters (look the help file, the 750 only has 5 CU) then see if the problem stop. I belive your configuration push to much the 750´s that´s could be the source of your problems.

Or you could start 2 instances of Boinc on the same host, 1 configurated to optimize the 780´s and the other for the 750´s.

Both will work but i prefear the 2 one since it will make your host optain the best of both worlds. In the first option your 780 will allways been underutilized.
ID: 1558692 · Report as offensive
Profile Mike Special Project $75 donor
Volunteer tester
Avatar

Send message
Joined: 17 Feb 01
Posts: 34253
Credit: 79,922,639
RAC: 80
Germany
Message 1558693 - Posted: 18 Aug 2014, 13:10:28 UTC

I cant say much about the drivers but i`ve heard they are causing trouble.

Your ffa values are defintely to high for 750ti.
With unroll 12 i would suggest max of 8192 4096.

Also you have AMD CPU`s are you reserving any CPU cores ?
Thats highly recommended.


With each crime and every kindness we birth our future.
ID: 1558693 · Report as offensive
Profile Zalster Special Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 27 May 99
Posts: 5517
Credit: 528,817,460
RAC: 242
United States
Message 1558696 - Posted: 18 Aug 2014, 13:18:17 UTC - in response to Message 1558693.  
Last modified: 18 Aug 2014, 13:23:33 UTC

Thanks, I'll change my configuration and roll back the driver. Only GPU crunching, no CPU. Setting for 0.5 CPU per Work units so total of 6 out of the 8 cores being used with 2 free.
ID: 1558696 · Report as offensive
Profile Cliff Harding
Volunteer tester
Avatar

Send message
Joined: 18 Aug 99
Posts: 1432
Credit: 110,967,840
RAC: 67
United States
Message 1558701 - Posted: 18 Aug 2014, 13:25:37 UTC - in response to Message 1558693.  

I cant say much about the drivers but i`ve heard they are causing trouble.

Your ffa values are defintely to high for 750ti.
With unroll 12 i would suggest max of 8192 4096.

Also you have AMD CPU`s are you reserving any CPU cores ?
Thats highly recommended.


I'm running with the 340.52 driver with my 2 x 750Ti with no problem. The only difference between your values and mine is the value for -unroll. Mine is set at 10 while your's is 12. I haven't had a chance to test the -tune option yet. http://setiathome.berkeley.edu/result.php?resultid=3679375027


I don't buy computers, I build them!!
ID: 1558701 · Report as offensive
Profile Zalster Special Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 27 May 99
Posts: 5517
Credit: 528,817,460
RAC: 242
United States
Message 1558703 - Posted: 18 Aug 2014, 13:30:16 UTC - in response to Message 1558701.  

I'm going to try it with the values Mike gave. If it doesn't freeze anymore then I'm good. If it does, then I'll look at rolling back the driver. My other system is similar to this one, only it has just 1-750. I haven't had any problems with freezing with that one, so I guess I'll wait and see how it progresses.

Thanks Mike, Juan and Cliff (hey Cliff, transplanted the guts of the old computer into that new Corsair case we talked about)


Zalster
ID: 1558703 · Report as offensive

Message boards : Number crunching : v0.42 with mixed GPUs


 
©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.