nvidia drivers crash a linux machine with 2 video cards

Question

In Fedora 11 (64bit) I was using the nvidia binary drivers (akmod-nvidia) to drive both by QUadro FX 1400, with 2 monitors on it, and my GeForce 8400 GX, with 1 monitor on it.

After upgrading to Fedora 12 I found that that setup would make the whole machine crash, it just hung, after about 5mins of use. The only way to fix this was to go to nvidia-settings and disable the FX 1400, that is, disable the third monitor.

I have not been able to find any reports of others with similar experience, but then again, it seems that no one is willing to put 2 video cards on a Linux machine.

As there any workarounds?

Answer 1

A lot of people run multiple gpu's on Linux boxes. I run a SLI set-up myself. First of all. Watch the temp of the gpu's. If they are getting hot that will do it. Often the fans on graphics cards will go bad without making any noise.

Next, I would try running the machine with only one or the other card. Swap them. And see if either one by itself causes the same problem.

Are you running compiz with all of those screens? Try disabling that.

When the system hangs, can you switch to a terminal? ctrl+alt+F1 If so, then your system is still working. Some software has locked up. Run ps -A and/or top to see if you can identify a process that could be causing the problem.

If the entire system is in fact locking up. Than my guess is that it is a more complex driver issue, or a hardware malfunction (overheating for example).

PS: For people having the (X wont start with two Nvidia cards) problem. If you do not have a BusID line for at least your primary videocard in your xorg.conf file that is the problem.

Answer 2

Try the free drivers? nv is pretty obfuscated w/ no 3d acceleration but I bet it supports XRandR better than the nvidia binary stuff (which only supports nvidia's proprietary Twinview software)

You're using the binary drivers. If nvidia doesn't support it...you have few options.

Answer 3

I'm having the same issue--except in my case, it's a single onboard nvidia display controller. The problem started with the basic installation, and has continued after installing the nvidia proprietary driver (190.53)...although it looks much better when it crashes. :-)

The display locks up, the mouse still works, but I can't ssh in from another machine, although the system is still on the network. If I have a running ssh from another machine, it still works--which indicates that something is causing the machine to become busy, not a complete lockup.

I'll try the CTL-ALT-F1 next time. But so far, ps, top, and logs haven't given me a clue--Xorg.0.log looks just fine, there arent' any dmesg indications, and the only "solution" is a hard reboot.

Funnily enough, this is my 8th FC12 install and only one machine is giving me headaches... A couple of the others are MythTv machines with nvidia cards that run just fine.