nV News Forums

 
 

nV News Forums (http://www.nvnews.net/vbulletin/index.php)
-   NVIDIA Linux (http://www.nvnews.net/vbulletin/forumdisplay.php?f=14)
-   -   1.0-9746, 1.0-9755 x86_64 system freeze with GeForce 8800GTX (http://www.nvnews.net/vbulletin/showthread.php?t=89233)

tachyon_john 04-04-07 01:02 AM

1.0-9746, 1.0-9755 x86_64 system freeze with GeForce 8800GTX
 
1 Attachment(s)
Hi,
We are running two identical Sun Ultra 40 workstations with GeForce 8800GTX cards installed in them. The systems previously had Quadro FX 3400 cards, and were quite stable. After installing the GeForce 8800GTX cards (they already had 1.0-9746 installed) we are now seeing system freezes if we load up VMD with a large molecule, start the molecule spinning, and then resize the window while it is animating. When the system freeze occurs, the machine is dead, unresponsive to ping etc, and cannot be recovered.

Since we're the developers of VMD, we have recently heard similar reports from two of our users at other sites, with different motherboards, different Linux distributions, and so on. Our users reported that the system lockups happened to them with both the 1.0-9746 drivers as well as 1.0-9755. We have only personally tested with 1.0-9746, but have verified the behavior they reported to us The problem appears to be specific to the 64-bit drivers/kernel, as one of our users reported that after reinstalling his system with a 32-bit Linux distribution, the crashes did not occur anymore. We thus believe the problem may be specific to both the GeForce 8800GTX and to 64-bit kernels. We have two CUDA test systems running 32-bit kernels with the 1.0-9751 and those machines do not appear to crash when tested in the same way, though we aren't doing much visualization on them since they are primarily intended as CUDA test machines.

I've attached the bug report log to this email. Let us know if you have suggestions or need help reproducing the problem. I looked through the driver bug reporting article on the forum and checked the obvious things with MMCONFIG etc, but didn't see anything that would explain this system lockup behavior we or our users are seeing with the 64-bit kernels and the 8800 cards with the latest drivers.

Cheers,
John Stone

Nordbryggan 04-04-07 01:59 AM

Re: 1.0-9746, 1.0-9755 x86_64 system freeze with GeForce 8800GTX
 
hi

Have you tried the new drivers, 1.0-9755?
They work like a charm on my home system,
680i, 8800GTX, etc, etc on opensuse 10.2 x64.
Very stable though the gfx cards do get a tad hot,
but i think thats normal for 8800.

Link:Nvidia x64 driver page

edit: Didn't notice that you already tried those, oopsa...

netllama 04-04-07 10:41 AM

Re: 1.0-9746, 1.0-9755 x86_64 system freeze with GeForce 8800GTX
 
John,
You stated that you heard reports of this problem also happening in other types of systems? Which systems were they?

Also, can you provide the dataset required to reproduce this?

thanks,
Lonni

tachyon_john 04-04-07 04:12 PM

Re: 1.0-9746, 1.0-9755 x86_64 system freeze with GeForce 8800GTX
 
Hi Lonni,
I've emailed the others that reported this problem to us and I've asked them to post to this thread with their system information or their corresponding nvidia-bug-report.log files.
I'll prepare a test case for you so that you can try to reproduce the problem. There didn't seem to be anything special about the dataset other than that it was a larger dataset.
We also caused the machine to lock while running another molecular graphics program (not written by us), so the problem occurs with more than just VMD as it turns out. I'll post a VMD test case for you tonight.

Thanks,
John

justinrocks 04-04-07 04:28 PM

Re: 1.0-9746, 1.0-9755 x86_64 system freeze with GeForce 8800GTX
 
Hi,

I've experienced the identical lockup on two different machines, identically configured, though different from John's setup. The machines are:

Dell Precision 390
Core 2 @ 2.4GHz
Driver: NVRM version: NVIDIA UNIX x86_64 Kernel Module 1.0-9755 Mon Feb 26 23:16:31 PST 2007
GCC version: gcc version 3.4.4 20050721 (Red Hat 3.4.4-2)

Card:
Model: GeForce 8800 GTS
IRQ: 161
Video BIOS: 60.80.0d.00.05

Kernel: 2.6.18.1-5smp, from CentOS release 4.2 (Final).

The lockup can be reliably reproduced in two different versions of VMD by loading a molecule (mol pdbload 1e79 from the VMD command line), switching to VDW representation, quickly rotating the molecule to produce a continuous animation, and then resizing the window.

Justin

tachyon_john 04-05-07 11:50 AM

Re: 1.0-9746, 1.0-9755 x86_64 system freeze with GeForce 8800GTX
 
Hi Lonni,
I've posted a test case script and datafile that you can use to recreate the problem on your x86_64 test system(s). The test data is here:
http://www.ks.uiuc.edu/Research/vmd/...8800bug.tar.gz

Cheers,
John Stone

stliston 04-05-07 02:11 PM

Re: 1.0-9746, 1.0-9755 x86_64 system freeze with GeForce 8800GTX
 
1 Attachment(s)
I am posting these nvidia-bug-report.log files at the request of tachyon_john who has been talking with my colleague Irvin here at the Univ. of Utah. Here are the details of the system we are running:

RHEL 4, x86_64 (2.6.9-42.0.10-smp kernel)
Single Intel Core 2 Duo(2.4GHz)
8GB RAM
Geforce 8800GTX with NVIDIA-Linux-x86_64-1.0-9755 driver.

Our bug is consistent the others: VMD (1.8.5 and 1.8.6b11); large molecule (~140,000 atoms); resize the window; total system lockup.

Attached are 2 log files zipped up: nvidia-bug-report.log1 in steady state while running VMD; nvidia-log-report.log2 in locked up state (normally this is not possible as the system is totally locked up, but while intentionally crashing things with trying to get these logs I found that after one lockup I still had an active ssh session and was about to run the script. Whether it will be of help I don't know).
Thanks,
Sam

netllama 04-05-07 08:22 PM

Re: 1.0-9746, 1.0-9755 x86_64 system freeze with GeForce 8800GTX
 
I'm able to reproduce this problem, however it appears to be specific to GeForce based G80 cards, and not Quadros. It should be fixed in the first released 100.xx series driver.

Thanks,
Lonni

tachyon_john 04-05-07 10:46 PM

Re: 1.0-9746, 1.0-9755 x86_64 system freeze with GeForce 8800GTX
 
Lonni,
Excellent! Thanks for tracking this problem down with us.

Cheers,
John Stone


All times are GMT -5. The time now is 10:13 PM.

Powered by vBulletin® Version 3.7.1
Copyright ©2000 - 2014, Jelsoft Enterprises Ltd.
Copyright 1998 - 2014, nV News.