CPU applications deployed and CUDA memory requirements changed

Message boards : News : CPU applications deployed and CUDA memory requirements changed
Message board moderation

To post messages, you must log in.

AuthorMessage
Profile Teemu Mannermaa
Project administrator
Project developer
Project tester

Send message
Joined: 20 Apr 11
Posts: 360
Credit: 758,716,620
RAC: 121,005
Message 477 - Posted: 26 May 2011, 12:41:04 UTC

New CPU applications deployed for 32-bit Windows and 64-bit Linux. Windows version should also run on 64-bit Windows systems. So you should now disable CPU applications on project preferences if you don't want to use your CPU for this project.

There's a known problem with checkpoint interval (uses a default 2h interval) and core selection (selected core is shared with ATI/nVidia on the same host) for these applications. Both problems should be fixed in next application version so you should wait for that if you have a problem due to either limitation.

Additionally, CUDA requirements were changed to accept cards with only 64MB of memory.
ID: 477 · Rating: 0 · rate: Rate + / Rate - Report as offensive
[SETI.USA]Tank_Master

Send message
Joined: 2 May 11
Posts: 4
Credit: 28,716,459
RAC: 0
Message 486 - Posted: 26 May 2011, 20:51:42 UTC
Last modified: 26 May 2011, 20:52:18 UTC

With the new nvidia memory requirnment, does http://moowrap.net/forum_thread.php?id=16 need updating? I updated http://boinc.berkeley.edu/wiki/GPU_computing#Attach_to_projects_with_GPU_applications pointing to it so users can get the most curent info.
ID: 486 · Rating: 0 · rate: Rate + / Rate - Report as offensive
[AF>FAH-Addict.net]toTOW

Send message
Joined: 4 May 11
Posts: 24
Credit: 39,957,464
RAC: 2,732
Message 488 - Posted: 26 May 2011, 21:43:51 UTC

Since you now have a CPU client, do you plan to add OGR support to Moo! Wrapper ?
ID: 488 · Rating: 0 · rate: Rate + / Rate - Report as offensive
Clod PatryProject donor
Volunteer moderator
Volunteer developer
Volunteer tester

Send message
Joined: 2 May 11
Posts: 65
Credit: 242,754,987
RAC: 0
Message 489 - Posted: 27 May 2011, 1:25:52 UTC

I changed http://boinc.berkeley.edu/wiki/GPU_computing#Attach_to_projects_with_GPU_applications, thanks for pointing.
Teemu will have to change the original post here.

I'm not sure we'll support OGR-27, since yoyo@home ( http://www.rechenkraft.net/yoyo/) is actually doing it.
Also, there's no GPU client for OGR at the moment.
ID: 489 · Rating: 0 · rate: Rate + / Rate - Report as offensive
JAMC

Send message
Joined: 19 May 11
Posts: 8
Credit: 4,742,137
RAC: 0
Message 492 - Posted: 27 May 2011, 4:10:36 UTC

A little heads up might have been nice as I came home to all 4 cpu cores cooking at 100% load in a 90'F room when I had CPU use set to disabled in my Moowrap preferences but then found it had been automagically enabled. Maybe default admin changes to not enable CPU and GPU use and let us enable to prevent possible meltdowns at our end.
ID: 492 · Rating: 0 · rate: Rate + / Rate - Report as offensive
Profile STE\/E
Avatar

Send message
Joined: 2 May 11
Posts: 57
Credit: 250,035,526
RAC: 0
Message 494 - Posted: 27 May 2011, 7:51:17 UTC - in response to Message 492.  

A little heads up might have been nice as I came home to all 4 cpu cores cooking at 100% load in a 90'F room when I had CPU use set to disabled in my Moowrap preferences but then found it had been automagically enabled. Maybe default admin changes to not enable CPU and GPU use and let us enable to prevent possible meltdowns at our end.


I agree, the Default should have been not to run the CPU Tasks & let those that want to run them Enable it themselves ...
ID: 494 · Rating: 0 · rate: Rate + / Rate - Report as offensive
Clod PatryProject donor
Volunteer moderator
Volunteer developer
Volunteer tester

Send message
Joined: 2 May 11
Posts: 65
Credit: 242,754,987
RAC: 0
Message 501 - Posted: 27 May 2011, 16:17:10 UTC

I also agree, CPU tasks should be disabled by default.
Like we already discussed, the CPU tasks won't make a huge impact, so i'm okay with disabling it by default.

However, the GPU utilization should be enabled by default.

Thanks for suggestion.
ID: 501 · Rating: 0 · rate: Rate + / Rate - Report as offensive
[AF>FAH-Addict.net]toTOW

Send message
Joined: 4 May 11
Posts: 24
Credit: 39,957,464
RAC: 2,732
Message 515 - Posted: 28 May 2011, 8:15:51 UTC - in response to Message 489.  
Last modified: 28 May 2011, 8:16:44 UTC

I'm not sure we'll support OGR-27, since yoyo@home ( http://www.rechenkraft.net/yoyo/) is actually doing it.
Also, there's no GPU client for OGR at the moment.


Your first argument is not valid since there are two wrappers for RC5 ;)

And since Moo! Wrapper is the only one allowing us to use our distributed.net account, I'd rather use it for OGR too instead of joining yoyo ...
ID: 515 · Rating: 0 · rate: Rate + / Rate - Report as offensive
Profile Teemu Mannermaa
Project administrator
Project developer
Project tester

Send message
Joined: 20 Apr 11
Posts: 360
Credit: 758,716,620
RAC: 121,005
Message 518 - Posted: 28 May 2011, 10:06:29 UTC - in response to Message 492.  

A little heads up might have been nice as I came home to all 4 cpu cores cooking at 100% load in a 90'F room when I had CPU use set to disabled in my Moowrap preferences but then found it had been automagically enabled.


I did try to warn about this change (that CPU apps are coming) in some post here, although, obviously didn't make it clear enough. Too late now, I think. :(

Odd that the CPU units got enabled by default. That's a default feature of BOINC server software and I didn't make any decision to enable or disable them by default. I would have thought that a previously disabled CPU would have stayed so.

Where did you set CPU disabled? We didn't have that setting in project preferences beforehand (it's not visible there unless we have one or more CPU apps deployed) so it must have been somewhere else.

I apologize from anybody who had problems due to me adding CPU applications.

-w
ID: 518 · Rating: 0 · rate: Rate + / Rate - Report as offensive
Profile Teemu Mannermaa
Project administrator
Project developer
Project tester

Send message
Joined: 20 Apr 11
Posts: 360
Credit: 758,716,620
RAC: 121,005
Message 519 - Posted: 28 May 2011, 10:08:58 UTC - in response to Message 486.  

With the new nvidia memory requirnment, does http://moowrap.net/forum_thread.php?id=16 need updating?


Latest requirements are documented in http://moowrap.net/forum_thread.php?id=41 post and that's kept updated as I make changes.

-w
ID: 519 · Rating: 0 · rate: Rate + / Rate - Report as offensive
vaio [The Lone Gunman]

Send message
Joined: 3 May 11
Posts: 41
Credit: 165,019,076
RAC: 0
Message 534 - Posted: 30 May 2011, 6:01:39 UTC
Last modified: 30 May 2011, 6:10:08 UTC

However I set up nvidia based rigs gpu usage is minimal....50% peaks and constant drops to practically zero.

This is even with 50% processors dedicated to feeding the gpu's.

Just checking if that is normal?

Addendum: Just changed core selection thingy fron -1 default to 1 and now getting a steady 50%....still rubbish but better than it was.

Anyone know the best setting for cuda please?
Team Renegades
Forum
ID: 534 · Rating: 0 · rate: Rate + / Rate - Report as offensive
vaio [The Lone Gunman]

Send message
Joined: 3 May 11
Posts: 41
Credit: 165,019,076
RAC: 0
Message 535 - Posted: 30 May 2011, 6:15:52 UTC - in response to Message 534.  
Last modified: 30 May 2011, 6:33:54 UTC

Darn.....it only worked for the one wu.....now gpu usage is all over the place again, fluctuating betwen zero and 50% constantly.

Addendum: Nevermind.

Sacrificing 2 cpu cores and get a max of 60% efficiency from my dual Cudas is pretty poor.

Relocated rig to target my other milestones....PG and Aqua.
Team Renegades
Forum
ID: 535 · Rating: 0 · rate: Rate + / Rate - Report as offensive
Profile Teemu Mannermaa
Project administrator
Project developer
Project tester

Send message
Joined: 20 Apr 11
Posts: 360
Credit: 758,716,620
RAC: 121,005
Message 563 - Posted: 1 Jun 2011, 19:29:22 UTC - in response to Message 534.  

Addendum: Just changed core selection thingy fron -1 default to 1 and now getting a steady 50%....still rubbish but better than it was.


For CUDA, best core value to use has been 9-11. You would need to run a benchmark (there's instructions on another post here) to find out for sure which is the best one on your host, though.

-w
ID: 563 · Rating: 0 · rate: Rate + / Rate - Report as offensive

Message boards : News : CPU applications deployed and CUDA memory requirements changed


 
Copyright © 2011-2017 Moo! Wrapper Project