Advanced search

Message boards : Graphics cards (GPUs) : Not only for fun... GeForce 8200 & GPUGrid...

Author Message
[B^S] BOINC-SG
Send message
Joined: 21 Apr 08
Posts: 2
Credit: 0
RAC: 0
Level

Scientific publications
wat
Message 3669 - Posted: 4 Nov 2008 | 23:34:53 UTC

I know its not really worth it, but at least I wanted to try it. Running GPUGrid ona GeForce 8200 OnBoard Chip...


05.11.2008 00:19:05||CUDA devices found
05.11.2008 00:19:05||Coprocessor: GeForce 8200 (1)
05.11.2008 00:19:05||Version change (6.2.19 -> 6.3.19)
05.11.2008 00:24:26|GPUGRID|Starting mLw6193-GPUTEST5-1-20-acemd_0
05.11.2008 00:24:26|GPUGRID|[task_debug] task_state=EXECUTING for mLw6193-GPUTEST5-1-20-acemd_0 from start
05.11.2008 00:24:26|GPUGRID|Starting task mLw6193-GPUTEST5-1-20-acemd_0 using acemd version 648
05.11.2008 00:24:27|GPUGRID|[task_debug] Process for mLw6193-GPUTEST5-1-20-acemd_0 exited
05.11.2008 00:24:27|GPUGRID|[task_debug] task_state=EXITED for mLw6193-GPUTEST5-1-20-acemd_0 from handle_exited_app
05.11.2008 00:24:27|GPUGRID|[task_debug] result state=COMPUTE_ERROR for mLw6193-GPUTEST5-1-20-acemd_0 from CS::report_result_error
05.11.2008 00:24:27|GPUGRID|[task_debug] Process for mLw6193-GPUTEST5-1-20-acemd_0 exited
05.11.2008 00:24:27|GPUGRID|[task_debug] exit code 1 (0x1): Unzulässige Funktion. (0x1)
05.11.2008 00:24:27|GPUGRID|Computation for task mLw6193-GPUTEST5-1-20-acemd_0 finished
05.11.2008 00:24:27|GPUGRID|Output file mLw6193-GPUTEST5-1-20-acemd_0_1 for task mLw6193-GPUTEST5-1-20-acemd_0 absent
05.11.2008 00:24:27|GPUGRID|Output file mLw6193-GPUTEST5-1-20-acemd_0_2 for task mLw6193-GPUTEST5-1-20-acemd_0 absent
05.11.2008 00:24:27|GPUGRID|Output file mLw6193-GPUTEST5-1-20-acemd_0_3 for task mLw6193-GPUTEST5-1-20-acemd_0 absent
05.11.2008 00:24:27|GPUGRID|[task_debug] result state=COMPUTE_ERROR for mLw6193-GPUTEST5-1-20-acemd_0 from CS::app_finished


Did I miss something? I exptected that it will work - not fast enough, but still... (??)

Installed NVidia 178.24, BOINC 6.3.19

I couldnt find the info if 8200 is CUDA 2.0 capable

Any advice is appreciated!

Cheers!
____________

Profile [BOINC@Poland]AiDec
Send message
Joined: 2 Sep 08
Posts: 53
Credit: 9,213,937
RAC: 0
Level
Ser
Scientific publications
watwatwatwatwatwat
Message 3670 - Posted: 5 Nov 2008 | 0:37:08 UTC - in response to Message 3669.
Last modified: 5 Nov 2008 | 0:38:06 UTC

Try 178.08

It have helped in very similar situation. (sry for bad english)
____________

Profile X-Files 27
Avatar
Send message
Joined: 11 Oct 08
Posts: 95
Credit: 68,023,693
RAC: 0
Level
Thr
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 3672 - Posted: 5 Nov 2008 | 1:56:56 UTC

Your error seems to be:
Cuda error in file 'deviceQuery.cu' in line 59 : out of memory.

Read more on this here:
http://www.gpugrid.net/forum_thread.php?id=340

Profile [SETI.USA]Tank_Master
Avatar
Send message
Joined: 8 Jul 07
Posts: 85
Credit: 67,463,387
RAC: 0
Level
Thr
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwat
Message 3673 - Posted: 5 Nov 2008 | 6:08:05 UTC

B^S - what OS are you running?

Profile UL1
Send message
Joined: 16 Sep 07
Posts: 56
Credit: 35,013,195
RAC: 0
Level
Val
Scientific publications
watwatwatwatwatwatwatwatwat
Message 3674 - Posted: 5 Nov 2008 | 6:18:53 UTC - in response to Message 3673.

B^S - what OS are you running?

Microsoft Windows XP
Professional x86 Editon, Service Pack 3, (05.01.2600.00)

Profile [SETI.USA]Tank_Master
Avatar
Send message
Joined: 8 Jul 07
Posts: 85
Credit: 67,463,387
RAC: 0
Level
Thr
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwat
Message 3677 - Posted: 5 Nov 2008 | 16:06:13 UTC

try vista or server 2008 (or win7 is you can get ahold of it) and see if the problem is still there. Im just curious.

Profile Krunchin-Keith [USA]
Avatar
Send message
Joined: 17 May 07
Posts: 512
Credit: 111,288,061
RAC: 0
Level
Cys
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 3678 - Posted: 5 Nov 2008 | 16:17:03 UTC - in response to Message 3669.
Last modified: 5 Nov 2008 | 16:40:41 UTC

I know its not really worth it, but at least I wanted to try it. Running GPUGrid ona GeForce 8200 OnBoard Chip...

Did I miss something? I exptected that it will work - not fast enough, but still... (??)

Installed NVidia 178.24, BOINC 6.3.19

178.24 is a CUDA 2.0 Driver, and under the driver release notes it lsits support for the 8200.

I couldnt find the info if 8200 is CUDA 2.0 capable

Go to http://www.nvidia.com/object/cuda_learn_products.html

It is listed as cuda capable

Any advice is appreciated!

Cheers!

[B^S] BOINC-SG
Send message
Joined: 21 Apr 08
Posts: 2
Credit: 0
RAC: 0
Level

Scientific publications
wat
Message 3679 - Posted: 5 Nov 2008 | 18:38:54 UTC - in response to Message 3677.

Thanx for the input. During the night, my computer froze. Since the only change since yesterday was to install Beta BOINC Client 6.3.19, I have no other idea what could have caused it... I downgraded to 6.2.19 again and will see if it happens again...

But thanx for your comments. Good to know that it should be possible, at least in theory on my host! :D

try vista or server 2008 (or win7 is you can get ahold of it) and see if the problem is still there. Im just curious.


I have none of these OSs... :( Sorry, can't satisfy your curiosity!

Profile [SETI.USA]Tank_Master
Avatar
Send message
Joined: 8 Jul 07
Posts: 85
Credit: 67,463,387
RAC: 0
Level
Thr
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwat
Message 3681 - Posted: 5 Nov 2008 | 19:42:18 UTC

you can download server 2008 form MS's site here:
http://www.microsoft.com/downloads/details.aspx?FamilyId=B6E99D4C-A40E-4FD2-A0F7-32212B520F50&displaylang=en
(at the bottom of the page)

ExtraTerrestrial Apes
Volunteer moderator
Volunteer tester
Avatar
Send message
Joined: 17 Aug 08
Posts: 2705
Credit: 1,311,122,549
RAC: 0
Level
Met
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 3687 - Posted: 5 Nov 2008 | 22:53:38 UTC

"Out of memory" .. the onboard GPUs usually don't have local memory and have to used the much slower system memory. It may be that GPU-Grid expects local memory, because crunching in the system mem is totally out of the question, performance wise.

In that case you could fiddle with drivers and OS as long as you want, it won't help ;)

MrS
____________
Scanning for our furry friends since Jan 2002

Scott Brown
Send message
Joined: 21 Oct 08
Posts: 144
Credit: 2,973,555
RAC: 0
Level
Ala
Scientific publications
watwatwatwatwatwat
Message 3689 - Posted: 5 Nov 2008 | 23:46:56 UTC

"Out of memory" .. the onboard GPUs usually don't have local memory and have to used the much slower system memory. It may be that GPU-Grid expects local memory, because crunching in the system mem is totally out of the question, performance wise.


I think that you are on the right track here, but I doubt it is the memory performance giving an "out of memory" error. Rather, since on-board GPUs frequently default to an "on-demand" memory allocation request of portions of system RAM, my guess would be that BOINC or the GPU app. doesn't like the way this request is handled. The original poster might check and see if a fixed allocation can be set to mimic a discrete card and memory.

ExtraTerrestrial Apes
Volunteer moderator
Volunteer tester
Avatar
Send message
Joined: 17 Aug 08
Posts: 2705
Credit: 1,311,122,549
RAC: 0
Level
Met
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 3699 - Posted: 6 Nov 2008 | 18:40:46 UTC

With older GPU-Grid clients if you were playing some relatively new game, GPU-Grid finished a WU in the background and BOINC tried to start up the next one, you'd get this "out of memory" error. Riva Tuner tells me that GPU-Grid uses less than 80 MB on the GPU, which shouldn't be a problem for 512 MB cards (also happened on GT200 cards with much more memory).

So what does this mean? The GPU could easily use system memory, if it runs out of local mem, so why this error? System memory is not full yet, the games are not that greedy. But historically they have been using one CPU core completely, irrespective of performance needs. I guess they're doing the same with local GPU memory, reserving space prior to actually needing it. In the past this would have been totally fine.

If I'm right with that, then I can explain the rest: the GPU-Grid app can distinguish between local and non-local memory, so it crashes when greedy games are running, despite the fact that there is still system memory available. And that means that the app also crashes with onboard GPUs, as they don't have local mem, regardless of what portion of system mem you tell them to use. Would be interesting to try though, any allocation >=128 MB *should* work.

(on a side note: I don't want to imply that this behaviour with GPU-Grid and games has changed in recent client versions, I just don't know)

MrS
____________
Scanning for our furry friends since Jan 2002

Post to thread

Message boards : Graphics cards (GPUs) : Not only for fun... GeForce 8200 & GPUGrid...

//