Linux driver issues 185

Message boards : Graphics cards (GPUs) : Linux driver issues 185
Message board moderation

To post messages, you must log in.

Previous · 1 · 2 · 3 · 4 · 5 · Next

AuthorMessage
movieman

Send message
Joined: 16 Jul 09
Posts: 3
Credit: 133,026,547
RAC: 0
Level
Cys
Scientific publications
watwat
Message 11174 - Posted: 18 Jul 2009, 12:53:10 UTC

Woke up this morning to find that the machine with the 260 card had errored out 3 WU..First at 6.5 hours and the other two in one -2 minutes..
The machine with the 9600 card is doing ok but showing almost 20 hours total to complete a WU..
ID: 11174 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
MarkJ
Volunteer moderator
Volunteer tester

Send message
Joined: 24 Dec 08
Posts: 738
Credit: 200,909,904
RAC: 0
Level
Leu
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 11175 - Posted: 18 Jul 2009, 13:09:04 UTC - in response to Message 11174.  

Woke up this morning to find that the machine with the 260 card had errored out 3 WU..First at 6.5 hours and the other two in one -2 minutes..
The machine with the 9600 card is doing ok but showing almost 20 hours total to complete a WU..


You are running the beta app aren't you? The "stock" app (acemd 6.64) we know doesn't work with later drivers.
BOINC blog
ID: 11175 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
movieman

Send message
Joined: 16 Jul 09
Posts: 3
Credit: 133,026,547
RAC: 0
Level
Cys
Scientific publications
watwat
Message 11176 - Posted: 18 Jul 2009, 14:56:54 UTC - in response to Message 11175.  

Woke up this morning to find that the machine with the 260 card had errored out 3 WU..First at 6.5 hours and the other two in one -2 minutes..
The machine with the 9600 card is doing ok but showing almost 20 hours total to complete a WU..


You are running the beta app aren't you? The "stock" app (acemd 6.64) we know doesn't work with later drivers.

Duh, uh no.. I'm running the "stock" app..
Ok, what driver should I be using with XP Pro 64 and a EVGA 260GTX?
Sign me,
Dave the clueless n00b..:)
ID: 11176 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
naja002
Avatar

Send message
Joined: 25 Sep 08
Posts: 111
Credit: 10,352,599
RAC: 0
Level
Pro
Scientific publications
watwatwatwatwatwatwatwat
Message 11178 - Posted: 18 Jul 2009, 15:56:01 UTC - in response to Message 11175.  

Woke up this morning to find that the machine with the 260 card had errored out 3 WU..First at 6.5 hours and the other two in one -2 minutes..
The machine with the 9600 card is doing ok but showing almost 20 hours total to complete a WU..


You are running the beta app aren't you? The "stock" app (acemd 6.64) we know doesn't work with later drivers.



I've been helping Movieman on another forum. He's running 6.6.36 and 186.18. Things seem to be working ok now, so it's wait-n-see atm. Hopefully everything will continue ticking away....
ID: 11178 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
SimonB

Send message
Joined: 25 Oct 08
Posts: 1
Credit: 2,258,127
RAC: 0
Level
Ala
Scientific publications
watwatwatwatwatwat
Message 11180 - Posted: 18 Jul 2009, 18:55:50 UTC

Hi Guys,

Looking for some advice re good Boinc/Driver combinations.

My current system is Win XP64, Boinc 6.6.28 and 186.18 drivers on an i7 with 3 GTX295's

The system runs one instance of GPUGRID (6.64) fine but subsequent units fail with a computation error almost straight away when they run on the other GPU's. All the GPU's are recognised by Boinc and work fine on Seti so I am a bit stumped with this problem.

Cheers, Simon.
ID: 11180 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
MarkJ
Volunteer moderator
Volunteer tester

Send message
Joined: 24 Dec 08
Posts: 738
Credit: 200,909,904
RAC: 0
Level
Leu
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 11186 - Posted: 19 Jul 2009, 5:09:08 UTC - in response to Message 11180.  

Hi Guys,

Looking for some advice re good Boinc/Driver combinations.

My current system is Win XP64, Boinc 6.6.28 and 186.18 drivers on an i7 with 3 GTX295's

The system runs one instance of GPUGRID (6.64) fine but subsequent units fail with a computation error almost straight away when they run on the other GPU's. All the GPU's are recognised by Boinc and work fine on Seti so I am a bit stumped with this problem.

Cheers, Simon.


I don't have any experience with XP64 as I run the 32 bit version of XP on my rigs. I run the 182.50 drivers which seem to be stable with GPUgrid.

You might want to upgrade your BOINC client to 6.6.36 or 6.6.37 as they did a couple of cuda fixes (to release video ram) in later versions, which may be part of your issues. I have 6.6.37 running on all my machines.

As for Seti, yes it seems a more forgiving cuda app (it was ported by nvidia). It doesn't seem to care what version drivers you run. GPUgrid are more particular, hence the attempt to get to a cuda 2.2 app that runs with "current" drivers.
BOINC blog
ID: 11186 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
JG

Send message
Joined: 28 Jun 09
Posts: 7
Credit: 0
RAC: 0
Level

Scientific publications
wat
Message 11250 - Posted: 22 Jul 2009, 12:50:12 UTC - in response to Message 11186.  

I am having problems with GPUGRID with drivers 185.18.08 in Linux. Is there anyway I could contribute to getting this issue fixed? I have experience developing CUDA programs and cross platform compiling.
ID: 11250 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile GDF
Volunteer moderator
Project administrator
Project developer
Project tester
Volunteer developer
Volunteer tester
Project scientist

Send message
Joined: 14 Mar 07
Posts: 1958
Credit: 629,356
RAC: 0
Level
Gly
Scientific publications
watwatwatwatwat
Message 11256 - Posted: 22 Jul 2009, 16:08:50 UTC - in response to Message 11250.  

nvidia is on it now.


gdf
ID: 11256 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
JG

Send message
Joined: 28 Jun 09
Posts: 7
Credit: 0
RAC: 0
Level

Scientific publications
wat
Message 11262 - Posted: 22 Jul 2009, 20:10:56 UTC - in response to Message 11256.  

So it is a problem with the drivers? Is there any information about what is happening with the driver that is causing this problem? It could be useful to know as I am having problems with one of CUDA programs and it could be related.
ID: 11262 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Michael Doerner

Send message
Joined: 28 Feb 09
Posts: 37
Credit: 666,889
RAC: 0
Level
Gly
Scientific publications
watwatwatwat
Message 11271 - Posted: 23 Jul 2009, 13:22:04 UTC - in response to Message 11262.  
Last modified: 23 Jul 2009, 13:27:10 UTC

Yes, the 185.X series does not work with GPUGrid. Revert to the 180.X series drivers until there's a fix.

Anyone tried the 190.X series yet? I'm going to try cranking out this last WU before I try it myself. Talking to the guys at forums.nvidia.com they said the 190.X initial release still had the bug, but I'll try it anyways just to see.

The frustrating part of all this is Nvidia initially kept blaming my configuration until someone else said they were having a problem. They never did get it through their thick skulls that the 180.X series work, and the 185.X series does not work regardless of the configuration. Only when I asked them what the results of THEIR OWN TESTING was did I get a responce that it was in fact a bug in the 185.X series, but they still blamed the GPUGrid programming as part of the problem.

Makes me long for and ATI card....Oh wait, not really......:-/

Mike D
ID: 11271 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile GDF
Volunteer moderator
Project administrator
Project developer
Project tester
Volunteer developer
Volunteer tester
Project scientist

Send message
Joined: 14 Mar 07
Posts: 1958
Credit: 629,356
RAC: 0
Level
Gly
Scientific publications
watwatwatwatwat
Message 11272 - Posted: 23 Jul 2009, 13:39:57 UTC - in response to Message 11271.  

It does not work for gpugrid because we are the only project on Linux.
Even substituting the application with a bogus cuda_malloc causes the problem.

They try to optimize resources, if it clear that there is a problem they will try to fix it. So, your posting on their forum is quite useful. Keep doing it.

Actually where you posted exactly? Let's write it here.:)

gdf
ID: 11272 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Michael Doerner

Send message
Joined: 28 Feb 09
Posts: 37
Credit: 666,889
RAC: 0
Level
Gly
Scientific publications
watwatwatwat
Message 11275 - Posted: 23 Jul 2009, 15:45:08 UTC - in response to Message 11272.  
Last modified: 23 Jul 2009, 15:45:54 UTC

Here's the post on the Nvidia forms where they admit there's a problem.....from "Netllama" @ Nvidia Corp. I keep asking about progress, but so far, they just keep releasing more crap. Also, in the reply is my original bug reports I posted on NVNews boards. Go ahead, pile on and pester them, maybe it'll get done quicker....

185.X Series doesn't work with GPUGrid Thread

Mike Doerner
ID: 11275 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Richard Haselgrove

Send message
Joined: 11 Jul 09
Posts: 1639
Credit: 10,159,968,649
RAC: 2
Level
Trp
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 11276 - Posted: 23 Jul 2009, 17:37:47 UTC

I didn't read this thread before I joined, so just went ahead and attached anyway. Had been running driver 185.85 so I could use the CUDA 2.2 DLLs (rt and fft) at SETI: no problems here at all.

Yesterday I upgraded to drivers 190.38, and CUDA 2.3 DLLs for SETI. Task 1006526 was run with the 190.38 drivers from start to finish.

Hosts are Windows XP SP3, 32-bit, with BOINC regularly upgraded to test new releases - I'm at v6.6.37 at the moment.

We had many reports of "driver errors" in the optimisation community at SETI - most of them were resolved when the hardware was returned under RMA. If the hardware and cooling are up to scratch, the new drivers work fine, even here.

Of course, that's not to say there aren't problems with the Linux drivers - but let's keep the two issues separate.
ID: 11276 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Michael Doerner

Send message
Joined: 28 Feb 09
Posts: 37
Credit: 666,889
RAC: 0
Level
Gly
Scientific publications
watwatwatwat
Message 11277 - Posted: 23 Jul 2009, 19:49:31 UTC - in response to Message 11276.  

What happened on your Linux box? That been the issue.

Mike Doerner
ID: 11277 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Michael Doerner

Send message
Joined: 28 Feb 09
Posts: 37
Credit: 666,889
RAC: 0
Level
Gly
Scientific publications
watwatwatwat
Message 11311 - Posted: 25 Jul 2009, 20:27:28 UTC - in response to Message 11277.  

Is the move to CUDA 2.2 going to work for Linux? The 185 and 190 series drivers do not work with GPUGrid on Linux. Are we just "out" until NVIDIA gets off their rear ends?

Mike D
ID: 11311 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile GDF
Volunteer moderator
Project administrator
Project developer
Project tester
Volunteer developer
Volunteer tester
Project scientist

Send message
Joined: 14 Mar 07
Posts: 1958
Credit: 629,356
RAC: 0
Level
Gly
Scientific publications
watwatwatwatwat
Message 11359 - Posted: 27 Jul 2009, 7:45:02 UTC - in response to Message 11311.  
Last modified: 27 Jul 2009, 7:45:22 UTC

You should install driver for CUDA2.1 as suggested in join page. Those work fine.
gdf
ID: 11359 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile GDF
Volunteer moderator
Project administrator
Project developer
Project tester
Volunteer developer
Volunteer tester
Project scientist

Send message
Joined: 14 Mar 07
Posts: 1958
Credit: 629,356
RAC: 0
Level
Gly
Scientific publications
watwatwatwatwat
Message 11380 - Posted: 27 Jul 2009, 16:37:46 UTC - in response to Message 11359.  

It seems that we were able to hack it. I have just upgraded a new Linux application. It's an incompatibility between the driver and the way boinc execute the application. Setting the PATH fixes it.

gdf
ID: 11380 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Michael Doerner

Send message
Joined: 28 Feb 09
Posts: 37
Credit: 666,889
RAC: 0
Level
Gly
Scientific publications
watwatwatwat
Message 11384 - Posted: 27 Jul 2009, 17:10:50 UTC

OK, here's an important update....I was able to get the 190.18 driver to start crunching successfully under openSUSE. Just not under any incarnation of the release openSUSE 11.1. I am now testing release 11.2 Milestone 4 from http://software.opensuse.org/developer. You will need to install the kernel packages listed on this openSUSE/NVIDIA page to get the 190.18 driver to sucessfully compile,however, I am now running the CUDA 2.3 version on openSUSE, so at least I won't be left behind after this weekend's swiutch.....;-).

Let me know if you have any other questions.

Mike Doerner
ID: 11384 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
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 11396 - Posted: 27 Jul 2009, 19:38:19 UTC - in response to Message 11384.  

Is your success related to the new app?

MrS
Scanning for our furry friends since Jan 2002
ID: 11396 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Michael Doerner

Send message
Joined: 28 Feb 09
Posts: 37
Credit: 666,889
RAC: 0
Level
Gly
Scientific publications
watwatwatwat
Message 11403 - Posted: 27 Jul 2009, 20:28:18 UTC - in response to Message 11396.  

No, I tried 190.18 on SUSE 11.1 before I decided to try Gentoo Linux. After that failed abysmally (yes, I'm a linux-lite person ;-) ), I tried Fedora 11, after I couldn't log into that installation as root, I downloaded openSUSE 11.2 Milestone 4 because I really didn't feel like reinstalling 11.1, which I knew was a problem, and then having to re-install KDE 4.2.4.

So far, my 9600 GSO is up to 14.573% and the task should be complete by tomorrow.

I would have really liked to try Fedora 11 a little more, but the restrictions through the GUI were killin' me.

Mike D
ID: 11403 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Previous · 1 · 2 · 3 · 4 · 5 · Next

Message boards : Graphics cards (GPUs) : Linux driver issues 185

©2026 Universitat Pompeu Fabra