Out of the blue, Cuda WU's start to use CPU, out of mem.

Questions and Answers : GPU applications : Out of the blue, Cuda WU's start to use CPU, out of mem.
Message board moderation

To post messages, you must log in.

AuthorMessage
Profile Gonad the Destroyer®©™
Avatar

Send message
Joined: 6 Aug 99
Posts: 204
Credit: 12,463,705
RAC: 0
United States
Message 884575 - Posted: 12 Apr 2009, 14:15:53 UTC

It started last night, not sure why, nothing has changed.
Any ideas what I can try???
I did enable SLi so only one card would be used, it looks like it might be using the card, but it, for lack of a better word, lags all the time, the timer to completion run's but the work seems to stop.

http://setiathome.berkeley.edu/result.php?resultid=1202126648

http://setiathome.berkeley.edu/result.php?resultid=1202124579

http://setiathome.berkeley.edu/result.php?resultid=1202124577

http://setiathome.berkeley.edu/result.php?resultid=1202123868
ID: 884575 · Report as offensive
Profile Jord
Volunteer tester
Avatar

Send message
Joined: 9 Jun 99
Posts: 15184
Credit: 4,362,181
RAC: 3
Netherlands
Message 884576 - Posted: 12 Apr 2009, 14:26:47 UTC - in response to Message 884575.  

If you see "Out of memory" errors, reboot the computer. There's a good chance that there's data stuck in the memory of the videocard and only a full power-cycle can clear that.
ID: 884576 · Report as offensive
Profile Gonad the Destroyer®©™
Avatar

Send message
Joined: 6 Aug 99
Posts: 204
Credit: 12,463,705
RAC: 0
United States
Message 884580 - Posted: 12 Apr 2009, 14:42:22 UTC
Last modified: 12 Apr 2009, 14:44:43 UTC

Shut down, restarted, something still isnt right.

The whole machine is lagged, but it doesnt seem to be using the CPU's.

I wonder if I didnt cook the GPU's, although they never got over 70°C


The WU's inm task man. are gobbling up each 100mb's ram, most of the CPU is system idel, but the machine is lagged out bad.....



The system is a Opteron 165 with 2gb's ram, the cards are 8800 GTS's 320mb, they are about 3 years old, but not abused to badly.....heh

EDIT:
I forgot to mention, there is a lot of HDD activity now.......
Like it;'s using the swap...
ID: 884580 · Report as offensive
Profile Jord
Volunteer tester
Avatar

Send message
Joined: 9 Jun 99
Posts: 15184
Credit: 4,362,181
RAC: 3
Netherlands
Message 884582 - Posted: 12 Apr 2009, 14:44:57 UTC - in response to Message 884580.  

Stop BOINC and see if that clears the problem. if you can stop BOINC, that is. I had a big slowdown of my system earlier caused by BOINC, just getting the Start button to open took 15 minutes. Logging off a further 30... ;-)
ID: 884582 · Report as offensive
Profile Byron S Goodgame
Volunteer tester
Avatar

Send message
Joined: 16 Jan 06
Posts: 1145
Credit: 3,936,993
RAC: 0
United States
Message 884584 - Posted: 12 Apr 2009, 14:56:50 UTC - in response to Message 884580.  

Shut down, restarted, something still isnt right.

The whole machine is lagged, but it doesnt seem to be using the CPU's.

I wonder if I didnt cook the GPU's, although they never got over 70°C


The WU's inm task man. are gobbling up each 100mb's ram, most of the CPU is system idel, but the machine is lagged out bad.....



The system is a Opteron 165 with 2gb's ram, the cards are 8800 GTS's 320mb, they are about 3 years old, but not abused to badly.....heh

EDIT:
I forgot to mention, there is a lot of HDD activity now.......
Like it;'s using the swap...


Hi,

Not sure if you're aware, but you also have several tasks showing "Completed, validation inconclusive" aside from the ones you posted.

ID: 884584 · Report as offensive
Profile Gonad the Destroyer®©™
Avatar

Send message
Joined: 6 Aug 99
Posts: 204
Credit: 12,463,705
RAC: 0
United States
Message 884623 - Posted: 12 Apr 2009, 16:34:17 UTC

Thanks for the heads up, I dont know what is going on here with this goofy thing....


But yes, when I pause BOINC the machine clears up and runs normal.

This is very annoying, and it just started happening, last night when my sister in law was surfing, oh crap, this reminds me, I wonder what the chances are it got infected....

I just reinstalled windows, and did have a virri scanner installed yet, and sister in law isnt all that smart, and she likes to check her emails here.......



I will check this right away....
ID: 884623 · Report as offensive
Profile Jord
Volunteer tester
Avatar

Send message
Joined: 9 Jun 99
Posts: 15184
Credit: 4,362,181
RAC: 3
Netherlands
Message 884627 - Posted: 12 Apr 2009, 16:41:14 UTC - in response to Message 884623.  

But yes, when I pause BOINC the machine clears up and runs normal.

VLAR (very low angle range) tasks can slow down when crunched on the GPU. You are aware of this, aren't you? (Else just check any of the other threads around, there'll be mention of it. ;-))

And no, there's no official way around it. Well, there is. The official way is to tell BOINC only to crunch on the GPU when the system is idle. The workaround is using the Lunatics team applications with the VLAR kill option.
ID: 884627 · Report as offensive
Profile Gonad the Destroyer®©™
Avatar

Send message
Joined: 6 Aug 99
Posts: 204
Credit: 12,463,705
RAC: 0
United States
Message 884637 - Posted: 12 Apr 2009, 16:55:26 UTC

Is there a quick way to check what the angle range is????


I will have a look around, I hope this is whats going on....
ID: 884637 · Report as offensive
Profile Jord
Volunteer tester
Avatar

Send message
Joined: 9 Jun 99
Posts: 15184
Credit: 4,362,181
RAC: 3
Netherlands
Message 884640 - Posted: 12 Apr 2009, 17:02:13 UTC - in response to Message 884637.  
Last modified: 12 Apr 2009, 17:04:47 UTC

Open the task (the 367KB file, no extension) in Notepad and look at what its value is for <true_angle_range>

I am not completely sure what value constitutes a low angle or a high angle or a normal angle, but I am sure someone will come along and tell you about that. Although, at a guess I would think anything below 0.01 is deemed VLAR.
ID: 884640 · Report as offensive
Profile Gonad the Destroyer®©™
Avatar

Send message
Joined: 6 Aug 99
Posts: 204
Credit: 12,463,705
RAC: 0
United States
Message 884653 - Posted: 12 Apr 2009, 17:51:05 UTC

I see what your saying now......

You think I should update the stock app with the Lunatics with the vlar killer in it??

I do recall trying before I got Cuda to run, and it had compute errors constantly..


I used Raistmer's_opt_pack_V10a_x86_SSE3_AMD_MB_CPU_GPU_team.rar, I have a lot of WU's, even though I told it to get .1 days worth...heh


I dont mean to be a pain in the butt, but what do I need to do to get this working???

The threads about it are a little confusing, and I dont wanna screw up a bunch of WU's again...




ID: 884653 · Report as offensive
Richard Haselgrove Project Donor
Volunteer tester

Send message
Joined: 4 Jul 99
Posts: 14650
Credit: 200,643,578
RAC: 874
United Kingdom
Message 884673 - Posted: 12 Apr 2009, 18:27:09 UTC - in response to Message 884640.  

... anything below 0.01 is deemed VLAR.

In theory they go up to 0.05, though you don't see many at 0.02 or above.
ID: 884673 · Report as offensive
Profile Jord
Volunteer tester
Avatar

Send message
Joined: 9 Jun 99
Posts: 15184
Credit: 4,362,181
RAC: 3
Netherlands
Message 884675 - Posted: 12 Apr 2009, 18:32:18 UTC - in response to Message 884653.  

I dont mean to be a pain in the butt, but what do I need to do to get this working???

If you don't mind, I will leave that to someone else to help you with. Perhaps it's even a good idea to ask about it in Number Crunching, as that forum's more read by the Lunatics (bar Richard and Byron).
ID: 884675 · Report as offensive
Profile Gonad the Destroyer®©™
Avatar

Send message
Joined: 6 Aug 99
Posts: 204
Credit: 12,463,705
RAC: 0
United States
Message 884682 - Posted: 12 Apr 2009, 18:43:49 UTC

Oh crap, I tried to get the Raistmer's_opt_pack working, and dang it, I lost a bunch......


ID: 884682 · Report as offensive

Questions and Answers : GPU applications : Out of the blue, Cuda WU's start to use CPU, out of mem.


 
©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.