Advanced search

Message boards : Graphics cards (GPUs) : GTX 285 fails on computation error

Author Message
mclaver
Send message
Joined: 9 Mar 09
Posts: 25
Credit: 694,215,882
RAC: 286,660
Level
Lys
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 21544 - Posted: 26 Jun 2011 | 16:39:21 UTC

All of my GPUGRID tasks are failing on computation error under Windows 7 profesional 64 bit, BOINC 6.12.26.

From the task

<core_client_version>6.12.26</core_client_version>
<![CDATA[
<message>
The system cannot find the path specified. (0x3) - exit code 3 (0x3)
</message>
<stderr_txt>
# Using device 0
# There is 1 device supporting CUDA
# Device 0: "GeForce GTX 285"
# Clock rate: 1.58 GHz
# Total amount of global memory: 1017839616 bytes
# Number of multiprocessors: 30
# Number of cores: 240
MDIO: cannot open file "restart.coor"
SWAN: FATAL : swanMemcpyDtoH failed

Assertion failed: 0, file swanlib_nv.c, line 390

This application has requested the Runtime to terminate it in an unusual way.
Please contact the application's support team for more information.

</stderr_txt>
]]>

From BOINC

6/26/2011 11:22:08 AM | GPUGRID | Computation for task A418-TONI_AGGsoup1-16-100-RND7971_0 finished
6/26/2011 11:22:08 AM | GPUGRID | Output file A418-TONI_AGGsoup1-16-100-RND7971_0_1 for task A418-TONI_AGGsoup1-16-100-RND7971_0 absent
6/26/2011 11:22:08 AM | GPUGRID | Output file A418-TONI_AGGsoup1-16-100-RND7971_0_2 for task A418-TONI_AGGsoup1-16-100-RND7971_0 absent
6/26/2011 11:22:08 AM | GPUGRID | Output file A418-TONI_AGGsoup1-16-100-RND7971_0_3 for task A418-TONI_AGGsoup1-16-100-RND7971_0 absent

____________

Paul Raney
Send message
Joined: 26 Dec 10
Posts: 115
Credit: 345,237,796
RAC: 0
Level
Asp
Scientific publications
watwatwatwatwatwatwat
Message 21545 - Posted: 27 Jun 2011 | 11:00:27 UTC - in response to Message 21544.

Is your card overclocked at all? The reference clock rate appears to be a little lower than what is reported in output.

It took a while to find overclocking settings that would work with my GTX 570 so you may want to start at reference clock speeds.

Paul Raney
Send message
Joined: 26 Dec 10
Posts: 115
Credit: 345,237,796
RAC: 0
Level
Asp
Scientific publications
watwatwatwatwatwatwat
Message 21546 - Posted: 27 Jun 2011 | 11:34:52 UTC - in response to Message 21544.

Please make sure you are setup in the GPUGrid preferences for shorter tasks only. Your successful tasks are all the shorter type. The GTX 285 is no longer recommended for longer tasks may not return results in time to get credit.

I looked at your computer and you are running the newest drivers. Have you tested the GPU configuration on other projects?

Please let us know if you get the issues resolved and how.

mclaver
Send message
Joined: 9 Mar 09
Posts: 25
Credit: 694,215,882
RAC: 286,660
Level
Lys
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 21570 - Posted: 1 Jul 2011 | 22:36:45 UTC - in response to Message 21546.

I do not overclock and Collatz and SETI seem to work fine. Right now I stopped running GPU Grid on this card. I am running GPUGRID on a GTX 275 and it has never had an error so I do not understand the comment that the GTX 285 is too slow.

Profile Carlesa25
Avatar
Send message
Joined: 13 Nov 10
Posts: 300
Credit: 67,023,334
RAC: 58
Level
Thr
Scientific publications
watwatwatwatwatwatwatwatwatwat
Message 21572 - Posted: 2 Jul 2011 | 11:43:51 UTC - in response to Message 21546.
Last modified: 2 Jul 2011 | 11:45:31 UTC

The GTX 285 is no longer recommended for longer tasks may not return results in time to get credit.


Hi, The GTX285 is a perfectly valid card to perform any task of this project, including long.

I am working with a GTX295 = x2 GTX285 downclock, are actually longer and the task ends in 15 hours and without activating SWAN_SYNC = 0. Greetings.

mclaver
Send message
Joined: 9 Mar 09
Posts: 25
Credit: 694,215,882
RAC: 286,660
Level
Lys
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 21573 - Posted: 2 Jul 2011 | 18:02:30 UTC - in response to Message 21572.

The GTX 285 is no longer recommended for longer tasks may not return results in time to get credit.


Hi, The GTX285 is a perfectly valid card to perform any task of this project, including long.

I am working with a GTX295 = x2 GTX285 downclock, are actually longer and the task ends in 15 hours and without activating SWAN_SYNC = 0. Greetings.



I have not done anything with SWAN_SYNC = 0. How do I check that and what do I need to do with it?

mclaver
Send message
Joined: 9 Mar 09
Posts: 25
Credit: 694,215,882
RAC: 286,660
Level
Lys
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 21574 - Posted: 2 Jul 2011 | 18:08:05 UTC - in response to Message 21573.

It looks like now I am getting 50% erorrs on Collatz. I have deinstalled and reinstalled the latest NVIDA drivers.

Driver Version 27533, Boinc 6.12.26, Windows 7 Profesional 64 biy.

Profile Carlesa25
Avatar
Send message
Joined: 13 Nov 10
Posts: 300
Credit: 67,023,334
RAC: 58
Level
Thr
Scientific publications
watwatwatwatwatwatwatwatwatwat
Message 21575 - Posted: 2 Jul 2011 | 18:52:14 UTC - in response to Message 21573.


I have not done anything with SWAN_SYNC = 0. How do I check that and what do I need to do with it?


Hello: Using SWAN_SYNC=0 is to dedicate one CPU core each core GPU with what results are achieved 10 to 15% better but at the expense of strong copper CPU load, I personally do not recommend it.

It has been discussed in several threads on this website, see: http://www.gpugrid.net/forum_thread.php?id=2123 and http://www.gpugrid.net/forum_thread.php?id=2553 and others.

On the other hand I see this using the GTX285 with 7% overclock, I think it's best not to push and set the nominal rate or a little less even to improve stability. Greetings.

mclaver
Send message
Joined: 9 Mar 09
Posts: 25
Credit: 694,215,882
RAC: 286,660
Level
Lys
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 21577 - Posted: 2 Jul 2011 | 21:58:29 UTC - in response to Message 21575.


I have not done anything with SWAN_SYNC = 0. How do I check that and what do I need to do with it?


Hello: Using SWAN_SYNC=0 is to dedicate one CPU core each core GPU with what results are achieved 10 to 15% better but at the expense of strong copper CPU load, I personally do not recommend it.

It has been discussed in several threads on this website, see: http://www.gpugrid.net/forum_thread.php?id=2123 and http://www.gpugrid.net/forum_thread.php?id=2553 and others.

On the other hand I see this using the GTX285 with 7% overclock, I think it's best not to push and set the nominal rate or a little less even to improve stability. Greetings.


That is not what I want to do. I have all my CPUs dedicated to WCG, and most of my GPUs dedicated to GPUGRID except this one.

I am actually ranked 19 in GPUGRID, and have GPUs from a GT 240 to multiple GTX 480s. The GTX 285 is the only one I am having problems with.

I am not that familiar with GPUs. I did not overclock this card, but I believe it was factory overclocked to 702, and that could be causing the problem.

How do I change the clock on a GPU (I know how to do it on a CPU), and what speed would you recommend for a GTX 285.

Dagorath
Send message
Joined: 16 Mar 11
Posts: 509
Credit: 179,005,236
RAC: 0
Level
Ile
Scientific publications
watwatwatwatwatwatwatwatwat
Message 21578 - Posted: 2 Jul 2011 | 22:12:19 UTC
Last modified: 2 Jul 2011 | 22:16:24 UTC

I have heard some Windows users are using MSI Afterburner to adjust clock speeds. I don't use Windows so I can't tell you much more than that. But before you mess with clock speeds, have you looked at the temperature your GPU is running at? It was running fine before, right? Maybe it just needs to have the dust blown out of it?

It seems every summer there is always a surge in errors on hardware that was running fine. It's usually just due to dust building up over the winter.

Profile skgiven
Volunteer moderator
Project tester
Volunteer tester
Avatar
Send message
Joined: 23 Apr 09
Posts: 3500
Credit: 869,046,807
RAC: 1,419,766
Level
Glu
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 21579 - Posted: 3 Jul 2011 | 0:00:19 UTC - in response to Message 21578.

I would go along with what Dagorath says:
Check for dust; in the summer it's warmer so any dust build up often causes heat problems. While this may be seen from products such as MSI Afterburner, EVGA Precision and GPUz, it might not always be so apparent; these products only report GPU temperatures, not capacitor temps.

If you feel the need to reduce speeds, drop the GPU clock speed. Your GTX285 is a CC1.3 card, so you may not even see a drop in performance. Keep the shader frequency as is. For CC2.0 and CC2.1 cards the GPU is linked to the shaders (1:2) so you would see a drop.

mclaver
Send message
Joined: 9 Mar 09
Posts: 25
Credit: 694,215,882
RAC: 286,660
Level
Lys
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 21582 - Posted: 3 Jul 2011 | 13:38:55 UTC - in response to Message 21579.

The fan is clean and the card does not seem to hot. I reduced my GPU clock from 702 to 650 and I am still getting about 30% errors on Collatz. Once I get through the queue for Collatz I will go back to GPU grid and see how I do there at the lower clock speed.

Profile skgiven
Volunteer moderator
Project tester
Volunteer tester
Avatar
Send message
Joined: 23 Apr 09
Posts: 3500
Credit: 869,046,807
RAC: 1,419,766
Level
Glu
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 21593 - Posted: 4 Jul 2011 | 17:35:44 UTC - in response to Message 21582.

30% is better than 50%, so you might be on to something.
I would suggest reducing the RAM frequency; it can often lead to better GPU stability.

Mican
Send message
Joined: 1 Jan 09
Posts: 2
Credit: 32,768,209
RAC: 76
Level
Val
Scientific publications
watwatwatwatwatwatwatwatwat
Message 21651 - Posted: 9 Jul 2011 | 8:19:19 UTC

I have the same error 0x3 message on every unit I tried this week and I am not conviced that is caused by overclocking or computer instability (this computer finished many tasks recently). Tasks are failing 5 secs after start, reporting missing or inaccessible file. Maybe is something wrong with a 275.33 nvidia driver?

<core_client_version>6.10.18</core_client_version>
<![CDATA[
<message>
Syst�m nem��e nal�zt uvedenou cestu. (0x3) - exit code 3 (0x3)
</message>
<stderr_txt>
# Using device 0
# There is 1 device supporting CUDA
# Device 0: "GeForce GTX 560 Ti"
# Clock rate: 1.90 GHz
# Total amount of global memory: 1008271360 bytes
# Number of multiprocessors: 8
# Number of cores: 64
MDIO: cannot open file "restart.coor"
SWAN: FATAL : swanMemcpyDtoH failed

Assertion failed: 0, file swanlib_nv.c, line 390

This application has requested the Runtime to terminate it in an unusual way.
Please contact the application's support team for more information.

</stderr_txt>
]]>

Profile skgiven
Volunteer moderator
Project tester
Volunteer tester
Avatar
Send message
Joined: 23 Apr 09
Posts: 3500
Credit: 869,046,807
RAC: 1,419,766
Level
Glu
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 21654 - Posted: 9 Jul 2011 | 12:01:26 UTC - in response to Message 21651.

I would suggest you install the driver again.
If that does not work reset the project from Boinc Manager.
You don't have to use SWAN_SYNC anymore.

[AF] Profanateur
Avatar
Send message
Joined: 25 Oct 08
Posts: 42
Credit: 42,812,268
RAC: 0
Level
Val
Scientific publications
watwatwatwatwatwatwatwatwatwatwat
Message 21672 - Posted: 12 Jul 2011 | 15:07:27 UTC

Same card and same problem on my only host wich have 570 GTX and the 285.
And no O/C on the 285.

There is few weeks the system work good, but now all wu on 285 go in errors absent.

And my temp are low. its not the problem :/

Hona
Send message
Joined: 21 Sep 10
Posts: 1
Credit: 130,680,273
RAC: 175,718
Level
Cys
Scientific publications
watwatwatwatwatwatwatwatwat
Message 22425 - Posted: 30 Oct 2011 | 13:09:52 UTC - in response to Message 21593.

This seems to be the way to go.
After some weeks of testing, my GTX 285 now runs fine with GPU and Shaders at factory-settings of 648 MHz and 1476 MHz. I only lowered the mem Frequency from 1242 MHz to 999 MHz and there seems to be no loss of speed in calculation for the rate of GPU-using is still the same.
No more compute- or validate-errors here and at Einstein@Home.
Thaks for your suggestion.
Hona

davidalary
Send message
Joined: 16 Nov 08
Posts: 4
Credit: 2,933,700
RAC: 0
Level
Ala
Scientific publications
watwatwat
Message 22493 - Posted: 9 Nov 2011 | 15:11:48 UTC - in response to Message 21651.
Last modified: 9 Nov 2011 | 15:17:35 UTC

I have the same error message. All my work unit on GPUGRID, Einstein@home, Seti using my GPU are a computation error. Is your problem was solved by installing again the graphics card drivers?

Let me know.

My error message:

<core_client_version>6.13.1</core_client_version>
<![CDATA[
<message>
Le chemin d�acc�s sp�cifi� est introuvable. (0x3) - exit code 3 (0x3)
</message>
<stderr_txt>
# Using device 0
# There is 1 device supporting CUDA
# Device 0: "GeForce GTX 560 Ti"
# Clock rate: 1.80 GHz
# Total amount of global memory: 1073741824 bytes
# Number of multiprocessors: 8
# Number of cores: 64
MDIO: cannot open file "restart.coor"
SWAN: FATAL : swanMemcpyDtoH failed

Assertion failed: 0, file swanlib_nv.c, line 390

This application has requested the Runtime to terminate it in an unusual way.
Please contact the application's support team for more information.

</stderr_txt>
]]>

Profile skgiven
Volunteer moderator
Project tester
Volunteer tester
Avatar
Send message
Joined: 23 Apr 09
Posts: 3500
Credit: 869,046,807
RAC: 1,419,766
Level
Glu
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 22494 - Posted: 9 Nov 2011 | 17:27:52 UTC - in response to Message 22493.

Try a driver reinstall.

Your Boinc configuration could influence the chances of failures, as could your environment (temperature), and what else you use the system for.
____________
FAQ's

HOW TO:
- Opt out of Beta Tests
- Ask for Help

davidalary
Send message
Joined: 16 Nov 08
Posts: 4
Credit: 2,933,700
RAC: 0
Level
Ala
Scientific publications
watwatwat
Message 22496 - Posted: 10 Nov 2011 | 14:38:13 UTC - in response to Message 22494.

I installed again all my graphic drivers card and I have always computation errors.

I have this kind of errors since I installed my new GTX 560TI graphic card. Before I used a 9800 GT and I don't have any problem.

Any one have a solution?

http://einstein.phys.uwm.edu/result.php?resultid=256463235

davidalary
Send message
Joined: 16 Nov 08
Posts: 4
Credit: 2,933,700
RAC: 0
Level
Ala
Scientific publications
watwatwat
Message 22498 - Posted: 10 Nov 2011 | 14:45:21 UTC - in response to Message 22494.

I installed again all my graphic drivers card and I have always computation errors.

I have this kind of errors since I installed my new GTX 560TI graphic card. Before I used a 9800 GT and I don't have any problem.

Any one have a solution?

Errors computation on my Boinc project

http://www.gpugrid.net/result.php?resultid=4525101

http://einstein.phys.uwm.edu/result.php?resultid=256463235

Dagorath
Send message
Joined: 16 Mar 11
Posts: 509
Credit: 179,005,236
RAC: 0
Level
Ile
Scientific publications
watwatwatwatwatwatwatwatwat
Message 22505 - Posted: 11 Nov 2011 | 0:30:31 UTC - in response to Message 22493.

davidalary wrote:
My error message:

<core_client_version>6.13.1</core_client_version>


I think we are not supposed to use any of the 6.13.xx clients because it causes problems creating WUs.

Post to thread

Message boards : Graphics cards (GPUs) : GTX 285 fails on computation error