Advanced search

Message boards : Number crunching : Other type of Calculations error

Author Message
Profile xmal
Avatar
Send message
Joined: 2 Feb 09
Posts: 8
Credit: 36,240,649
RAC: 0
Level
Val
Scientific publications
wat
Message 6576 - Posted: 11 Feb 2009 | 20:42:44 UTC

Hello! I have several WU stoping immediatly after starting with calculation error. Until today everything was ok, crunching without probleme. But Today, all new WU are in calculation error. When I look the workid status on the site here is the reason:
<core_client_version>6.4.5</core_client_version>
<![CDATA[
<message>
process exited with code 1 (0x1, -255)
</message>
<stderr_txt>
# Using CUDA device 0
Cuda error in file 'deviceQuery.cu' in line 59 : out of memory.

The task id example was: Task ID 299733

Is there something I could do ? What ? Actually I'm far from being short in memory:
Memory: 1.4Gio/2.0 (71.9%)
Swap: 591.6Mio/3.0Gio (19.2%)
Thanx in advance for any help! (reset the project does nothing).

Profile Paul D. Buck
Send message
Joined: 9 Jun 08
Posts: 1050
Credit: 37,321,185
RAC: 0
Level
Val
Scientific publications
watwatwatwatwatwatwatwatwatwat
Message 6593 - Posted: 12 Feb 2009 | 11:12:02 UTC

The problem is the amount of memory on the GPU card. You have 255MB on the 8600GT ... and that may not be enough for the models.

I wonder if the server needs to be smarter about which models it assigns out to the GPUs ...

Profile xmal
Avatar
Send message
Joined: 2 Feb 09
Posts: 8
Credit: 36,240,649
RAC: 0
Level
Val
Scientific publications
wat
Message 6598 - Posted: 12 Feb 2009 | 16:45:01 UTC - in response to Message 6593.

The problem is the amount of memory on the GPU card. You have 255MB on the 8600GT ... and that may not be enough for the models.


But the Wu just before had no trouble, is there so big differences between WU ? And I've tryed about 10 WU after with the same problem!


I wonder if the server needs to be smarter about which models it assigns out to the GPUs ...


On my end is there something I could do to free or reorganize the memory of my card ?

Profile Paul D. Buck
Send message
Joined: 9 Jun 08
Posts: 1050
Credit: 37,321,185
RAC: 0
Level
Val
Scientific publications
watwatwatwatwatwatwatwatwatwat
Message 6599 - Posted: 12 Feb 2009 | 17:16:14 UTC

There are at least three different "sizes" of task as far as run time. I would guess that one or more of those tasks takes up more memory than others. Even more likely is the slight differences between various models will use more or less memory than others.

Sadly, GPU cards are fixed in size as to the amount of memory that they contain.

You COULD try to turn off visual effects, use a single color background to see if that makes a difference, though I would not hold out much hope ...

This is one of the reasons that long ago I settled on getting cards with at least 1G of VRAM ... I kept running into issues with some games back when I was a more active gamer.

In that I am just another participant like you, I cannot advise you better ... perhaps ETA or one of the other project types could stick an oar in and suggest something ...

Profile xmal
Avatar
Send message
Joined: 2 Feb 09
Posts: 8
Credit: 36,240,649
RAC: 0
Level
Val
Scientific publications
wat
Message 6600 - Posted: 12 Feb 2009 | 18:06:51 UTC - in response to Message 6599.

There are at least three different "sizes" of task as far as run time. I
[...]
Sadly, GPU cards are fixed in size as to the amount of memory that they contain.
[...]
if that makes a difference, though I would not hold out much hope ...


Ok ok, I understand, thanx for your answer.


This is one of the reasons that long ago I settled on getting cards with at least 1G of VRAM ... I kept running into issues with some games back when I was a more active gamer.


Well, I should have meet you before buying mine... :-/ Number of shade is also important but at the end this makes a very expensive card no ?


In that I am just another participant like you, I cannot advise you better ... perhaps ETA or one of the other project types could stick an oar in and suggest something ...


ETA ??? oar ? I don't understand these words, sorry.

Anyway, thank you very much!

Profile Paul D. Buck
Send message
Joined: 9 Jun 08
Posts: 1050
Credit: 37,321,185
RAC: 0
Level
Val
Scientific publications
watwatwatwatwatwatwatwatwatwat
Message 6604 - Posted: 12 Feb 2009 | 20:29:44 UTC

ETA: ExtraTerrestrial Apes

He is one of the forum moderators ... there are some other project people and most of the time they know way more than I ... but sometimes I have a little clue and make suggestions to save them time ... then, I run out at times and have to hope they might have something more ...

"Stick the oar in..." is an English language (perhaps only American) expression meaning that they might add to what I said... kinda a polite "butt into the conversation" hint ... and an oar is the thing you row a boat / canoe with ...

Higher capabilities do imply higher costs ... but, to my mind, less expensive in the long run in frustration if nothing else. There are a number of cards that are adequate to good on GPU Grid including my lowest end card a 9800 GT which cost me $125 when I bought it a little while ago. The more capable cards I have bought are indeed up there in price at about $400 to $500 each ...

On the other hand, I am a BOINC addict and it seems a low enough price to me ...

Profile Stefan Ledwina
Avatar
Send message
Joined: 16 Jul 07
Posts: 464
Credit: 221,007,857
RAC: 4,333,521
Level
Leu
Scientific publications
watwatwatwatwatwatwatwat
Message 6612 - Posted: 13 Feb 2009 | 16:53:09 UTC - in response to Message 6576.

Hello! I have several WU stoping immediatly after starting with calculation error. Until today everything was ok, crunching without probleme. But Today, all new WU are in calculation error. When I look the workid status on the site here is the reason:
<core_client_version>6.4.5</core_client_version>
<![CDATA[
<message>
process exited with code 1 (0x1, -255)
</message>
<stderr_txt>
# Using CUDA device 0
Cuda error in file 'deviceQuery.cu' in line 59 : out of memory.

The task id example was: Task ID 299733

Is there something I could do ? What ? Actually I'm far from being short in memory:
Memory: 1.4Gio/2.0 (71.9%)
Swap: 591.6Mio/3.0Gio (19.2%)
Thanx in advance for any help! (reset the project does nothing).


What driver are you using? One from Nvidia or a driver of your Distro?
I'm using 177.82 with my 9800GTX and 8800GT on Ubuntu without much problems...

____________

pixelicious.at - my little photoblog

Profile xmal
Avatar
Send message
Joined: 2 Feb 09
Posts: 8
Credit: 36,240,649
RAC: 0
Level
Val
Scientific publications
wat
Message 6791 - Posted: 19 Feb 2009 | 21:03:03 UTC - in response to Message 6604.
Last modified: 19 Feb 2009 | 21:04:58 UTC

...

Post to thread

Message boards : Number crunching : Other type of Calculations error

//