Is your linux freezing with an NVIDIA video card?
 

I'm having the exact problem in the title. After seven months of digging I've come to the conclusion that the problem is somewhere between kernel and X server. I've tried contacting several firends with guru-like experience, I tried looking for answers on FedoraForum, I asked a question there, and I even started a bug report where the blame keeps being passed around different components.

I thought the problem might be isolated to my system, as I wasn't able to find any concrete info on this. About a week ago, I finished a new build for a friend. The PC has an NVIDIA card (GTX460). I tried Fedora 13 on that computer and I noticed the problem there as well. Actually, it was more pronounced than on my system. You couldn't use the system (friend's build) for more than two hours.

I'd like to get a little feedback on just how spread this problem is. I'm trying to see if it's a problem with NVIDIA cards in general, with Fedora in particular, if it's confined to 64-bit systems, etc.

Whether or not you have this issue, please post about it. Please also post to indicate that you have no problem if that is the case. I'll start the first "report". Please state all the items in your report.

Video Card:
Dual GeForce9800GT

Driver:
nvidia (latest akmod-nvidia from rpmfusion) but problem also present with nouveau

Kernel:
Noticed problem from 2.6.32 (earliest tested) to 2.6.34 (latest tested). I did not test with earlier kernels

Problem:
System randomly freezes. In most cases, keyboard and mouse stop responding (CapsLock does not toggle the light on the keyboard). system stops responding to ssh and ping. A hard reboot is required
On rare occasions, the ssh login is possible and restarting the X server usually revives the system.

Have I lost my mind yet:
Yes

It's interesting that it happens with nouveau and nvidia. Have you tried a different distribution for kicks? I recommend Arch as most packages are minimally patched: so you can determine if it's an upstream issue or Fedora-specific. Also do you use Windows at all? Because it if happens there too, you might just have a faulty card.

Is the OP asking this question in a Fedora-specific context, or does OS not matter here?

Just in case you want comparisons from other OS's:

I currently have zero problems like that, using two nvidia 7200/7300 PCIE cards, the blob driver (260.xx.xx released like today) and Slackware64-current and kernel=2.6.35.x. I just upgraded to this new driver today, and it works great, even with MSI enabled.

However, not all has *always* been peachy. If you want to hear about problems in the past, with older X servers (within the past year) and previous blob drivers, maybe you'd enjoy this other thread:
http://www.linuxquestions.org/questi...rg-bug-794887/

Check your logs too - there's usually some hint of *something*, such as those "NVIDIA(0) WAIT blah blah" errors I used to get when my X hung as described in that thread.

EDIT: For the record & clarity, I should have also stated that with the exception of the particular combination of kernel, driver and X server that was giving me a hard time in the thread I linked to above, historically for me the nvidia driver(s) and hardware have worked pretty much perfectly.

I've been using nvidia cards from years and I've had no stability problems at all with them. Currently using a nvidia 7300 gt card on Debian Sid. Used a 7300 gs until a couple of months ago with no problems at all either. My kernel version is 2.6.32-5-bigmem (32 bits with PAE enabled, although I've used 64 bit systems on this machine before without any stability problems).
As Meson suggested, might be a distro-related problem, or maybe a faulty card... or something else. Have you checked the system logs when this happens?

edit: I forgot to say, I'm using the proprietary driver, version 195.36.31. I've also used the newer versions without any issue, but I switched to this one because it's the one on debian repositories.

Regards.

After following fedoraforums tutorial and blacklisting nouveau and installing the proprietary driver (shortly after installing fedora) everything has been smooth sailing. Kernel updates are no problem. I'm using 256.53 driver for my Nvidia Quadro FX 570M w/ 512MB DDR3 (essentially an 8600M GT) on this laptop.

I had problems with nvidia and ati video with recent kernels. The solution for me was to add the "nomodeset" kernel boot option.

zero issues on OpenSUSE 11.3 & CentOS 5.5 with a GTS 240
-- but i am using the driver in my sig and not a rpm --

and zero problems with a VERY old geforce 2 mx400 with Arch

Adding "nomodeset" was the trick here as well. Most distros with newer kernels have nouveau in the kernel, which causes problems with the Nvidia driver.

On Debian/Ubuntu system edit /etc/default/grub
GRUB_CMDLINE_LINUX_DEFAULT="nomodeset"

I have used nvidia forever, haven't had any issues that I can remember.

System: Host craigevil Kernel 2.6.35-6.dmz.3-liquorix-686 i686 (32 bit) Distro Debian GNU/Linux squeeze/sid
CPU: Single core Intel Pentium 4 (-HT-) cache 1024 KB flags (lm nx sse sse2 sse3) bmips 5585.93
Clock Speeds: (1) 2792.966 MHz (2) 2792.966 MHz
Graphics: Card nVidia G96 [GeForce 9400 GT] X.Org 1.7.7 Res: 1280x1024@50.0hz
GLX Renderer GeForce 9400 GT/PCI/SSE2 GLX Version 3.3.0 NVIDIA 260.19.06 Direct Rendering Yes

Thanks for all your replies.

@ Meson
Windows doesn't have this problem on the two systems I tested, but Fedora does.

@GrapefruiTgirl
I haven't tried other distributions. I remember briefly using Fedora on a system with a 7200 (or 7400) with no issues. I was actually hoping to find a correlation between the problem and 8000-series or higher cards The logs are fairly useless. I've had some people look over them, and nothing susicious. I even saved some logs from when the display froze, but ssh still worked., with the same degree of failure.

I have space for two more distributions, but testing that would take me weeks since the problem is very sporradic: sometimes it freezes in less than an hour, sometimes it goes on for days without a glitch. I'm really afraid of the effort of reconfiguring everything.

@jschiwal
Could you please elaborate on the problems you were having? I actually tried the nomedeset trick, and it didn't seem to change anything.

@all
Please continue to post about problems with nvidia cards or the lack of. The more info we have, the easier it will be to draw a conclusion as how wide (or narrow) spread this issue is. And then hopefully someone wiser will be able to get to the bottom of it.

In the mean time, I'll try test a different distribution (probably Debian or Ubuntu) and see if I can reproduce the issue there.

in general, i don't have that problem on my installs. However, if i do have a recurring odd problem related to the Nvidia driver, i usually revert to an older version, and see it still occures.

Is that a typo? It is nomodeset and not nomedeset.

I don't remember the particular problem I had with nvidia. It was a while ago. On my netbook with an older ATI mobile graphics card, there was a problem with horizontal jitter. Even at the login screen, I could see white streak flashes.

This was using the radeon driver.

Some versions of the nvidia driver do cause instability, but currently with kernel 2.6.35.7 and nvidia 256.53 (64-bit) everything is 100% stable so far.

Two machines, two cards, same package versions, two different experiences:

#1: 2x Opteron 8222, GT 430 PCIe - perfectly stable with binary driver, running various 2D, 3D and GP apps without issue.
#2: 2x Opteron 290, 7800 GS AGP - random hangs with binary driver, similar to OP's problem. some apps crash the machine with nouveau driver, others run fine. See my thread for details.

#3: going off topic, same machine as #2 but with an ATI Radeon HD 4850 AGP - binary driver bogged down and froze when scrolling or moving windows, radeon driver was stable for 2D but would crash when running most 3D apps.

I haven't ever had a problem with display freezing unless there was a misconfiguration someplace. I haven't had any problems at all with nVidia drivers that I can document for several years.

I presently DO have some as-yet unidentified problem in my system that is causing hard lockups at intermittent intervals (separated, commonly, by several days), but I am pretty sure at this point that it has something to do with my Windows 7 Professional installation running in VMWare 7.2 on my Mandriva system with a 2.6.33 32 bit kernel and PAE enabled. Whether the nVidia driver is implicated, I cannot say, but I don't think so.