Nvidia Linux 325.15 driver release

[Nicolas_orleans]

I just noticed Nvidia released today new certified 325.15 drivers http://www.nvidia.com/object/linux-disp ... river.html. However, based on changelog I am not sure it will solve the issues that appear after 24-36 hours folding.
Anyone tried them already ?
From my side I will wait for xorg-edgers to package them properly for Ubuntu 13.04

[Nicolas_orleans]

325.15 upgrade from xorg-edgers went fine under Ubuntu 13.04
Folding since 8 hours without issues
Will post later what happens after 36 hours...

[Nicolas_orleans]

Same behaviour (PPD drop) after 36 hours as for 319.32. Still fixed by rebooting.

[bruce]

Same behaviour (PPD drop) after 36 hours as for 319.32. Still fixed by rebooting.

Is a reboot required, or is it enough to Pause and resume folding? Does finishing a WU and getting a new one reset it sufficiently?

[bollix47]

Nvidia is still working on the problem but as of two days ago they did not have a fix yet.

Source: https://forums.geforce.com/default/topi ... 6/#3887136

This is a 36 hour TDR that affects some 700 series GPUs using 320.xx drivers whether they are working at load or sitting idle.

[Nicolas_orleans]

Is a reboot required, or is it enough to Pause and resume folding? Does finishing a WU and getting a new one reset it sufficiently?

It actually kills the OS GUI, so you can't access nor use FAH Control.

As bollix said, this Series 700 problem is well identified by NV, and they are working on it since beginning of July

[bruce]

Understood.

I don't see any such problem with my 650Ti so I'm just wondering how it fails for those of you with 700 GPUs. Here's hoping for a prompt fix.

[ChristianVirtual]

First it bite away your interrupts and keyboard/mouse don't work anymore. At that time network/ssh still works. But after some time even here sudo for a proper shutdown will fail too. Then only the hard reset/power cycle works. Quite nasty.

[pmsfh2008]

Looks like I have been getting this too with my new 770.
Twice now my system has locked up and required a reboot.
Running 320.49 & seems to occur with 7810 WU's.
Something in Chrome seems to initiate the lock up.
Yes, nasty.

[bruce]

Something in Chrome seems to initiate the lock up.
Yes, nasty.

I don't have a 700 to test nor do I have Chrome, but if it really is something in the browser, there's probably a setting to disable Chrome's use of the GPU. I know other browsers have an obscure setting that switches it back to CPU-based rendering.

[ChristianVirtual]

I don't run Chrome ... In few cases i run Firefox on Ubuntu. But really rare as the folding box is dedicated and dont require a browser. For me Chrome is not the reason. It looks like some mismanagement of system resources where (maybe) some race condition with the faster/newer cards cause an exhaustion of vital resources (as we can see with the blocked interrupts for keyboard and mouse). Still hoping that NV get it fixed soon.

[bollix47]

They've found the fix!!!

https://forums.geforce.com/default/topi ... 8/#3889648

[ChristianVirtual]

After weeks of daily reboot I forgot it once and bummm: the 36 hour bug bite me again ... Great ... Took my SMP:5 with in the reboot. Lucky the checkpoints recovered well after the unfriendly reset.
Seems NV didn't yet fixed the Linux driver.

[Nicolas_orleans]

Nvidia released today the 331.13 linux beta drivers. According to changelog, lot of bug fixes... but I did not see the 36 hours TDR fix...
I am consequently still on 325.15 with daily reboot

If anyone tries 331.13 beta , it would be nice to report here if TDR has been fixed and if yes, if folding performance has been affected (on Windows, 331.40 appears NOT to be recommended except for GTX 780 owners)