L4D2 crashes X.org on Linux (Nvidia)

[darkargon]

Starting the first level in single player crashes the X.org Server. The intro runs fine @ ~120FPS but after the game gives me the ability to control my char the game freezes and my audio starts to loop. After 1-2 Minutes i get weird rendering artifacts and finally my 2 screens turn black (no signal).

syslog & system information: https://gist.github.com/darkargon/1e98b533db2a7bd60934

Linux Kernel: 3.9.0 (also happes with 3.8)
Nvidia Driver: 319.17 (did also happen with older versions)

[Plagman]

Hmm, the syslog indicates a hardware problem; can you check the temperatures on your GPU core while you're playing and see if the problems correlate with a big increase? You can periodically append nvidia-settings -q GPUCoreTemp to a file or something along these lines.

[darkargon]

I don't think its a heat problem, other source games run fine (Portal for example).
Also i can't check the temperature while "playing" because as I said I can't even move, it instantly freezes after the intro.

L4D2 runs fine under Windows on this box.

[ZeXx86]

I was affected by GPU overheating only in CS:Source, all other games were without issue. Then I cleaned my gpu fan and all was OK. Temperature went down to 50 from 70°C

[darkargon]

This seems to be a compiz/unity issue. L4D2 runs well under gnome 3. No crashes.

[jwcalla]

I'm experiencing a similar issue pretty frequently in a campaign called Day Break: http://www.l4dmaps.com/details.php?file=27072 -- almost always when the safe house loads after playing through the first map. I've only seen it randomly in other campaigns.

The behavior is:

1. Displayed frame rate drops to about 1 frame every 4-5 seconds. The same sound bit loops over and over until the next frame is displayed.
2. hl2_linux process goes to ~180%
3. X.org process goes to 100%+.
4. GPU Utilization to 99% (ordinarily 30-40% in-game).
5. PCIe Bandwidth Utilization to 99% (ordinarily <10% in game).
6. The mouse cursor appears to alter between being grabbed and not grabbed, as I can move the cursor to compiz hot corners, my second display, etc. and interact (barely) with the desktop.
7. The user session is essentially crippled and I have to do a VT switch and log into a console session and kill the hl2_linux process to regain control.

Some things I've ruled out:

• no errors in kernel log, syslog, terminal console, game console, Xorg.log or .xsession-errors.
• gpu temps are normal
• single-monitor setup is no different
• turning off threaded optimizations was no different

I've tried drivers 331.17, 331.13, 325.15, and something from the 319 branch.
It seems to occur primarily when both AA and anisotropic filtering are on (max). I didn't see the issue when one or the other was off.
I didn't experience the problem in gnome-shell or compiz + gnome-panel. Only compiz + unity.

I'm wondering if it's VRAM-related. During play the VRAM is always maxed out. On Windows, VRAM usually starts around 400 MB at the beginning of a map and then works up to as high as 1100 MB by the end of the map, then resets to ~400 MB again in the safe house. I typically have Chromium running in the background with a crapton of tabs open so now I'm investigating if that is a contributing factor.

Some people have mentioned similar "hangs" in comments to other bug reports. E.g., #600, #1422.

Does Valve or NVIDIA have some kind of tracing program we could run while the game is hammering the CPU / GPU so that we can dig further to see what operations it's attempting to do?

Processor Information:
    Vendor:  GenuineIntel
    CPU Family:  0x6
    CPU Model:  0x1e
    CPU Stepping:  0x5
    CPU Type:  0x0
    Speed:  2927 Mhz
    8 logical processors
    4 physical processors
    HyperThreading:  Supported
    FCMOV:  Supported
    SSE2:  Supported
    SSE3:  Supported
    SSSE3:  Supported
    SSE4a:  Unsupported
    SSE41:  Supported
    SSE42:  Supported
    
Network Information:
    Network Speed:  
    
Operating System Version:
    Ubuntu 12.04.3 LTS (64 bit)
    Kernel Name:  Linux
    Kernel Version:  3.2.0-56-generic
    X Server Vendor:  The X.Org Foundation
    X Server Release:  11103000
    X Window Manager:  Compiz
    Steam Runtime Version:  steam-runtime-release_2013-10-23
    
Video Card:
    Driver:  NVIDIA Corporation GeForce GTX 570/PCIe/SSE2
    Driver Version:  4.4.0 NVIDIA 331.17
    OpenGL Version: 4.4
    Desktop Color Depth: 24 bits per pixel
    Monitor Refresh Rate: 60 Hz
    VendorID:  0x10de
    DeviceID:  0x1081
    Number of Monitors:  2
    Number of Logical Video Cards:  1
    Primary Display Resolution:  1920 x 1080
    Desktop Resolution: 3840 x 1080
    Primary Display Size: 20.91" x 11.77"  (23.98" diag)
                                            53.1cm x 29.9cm  (60.9cm diag)
    Primary Bus: PCI Express 16x
    Primary VRAM: 1280 MB
    Supported MSAA Modes:  2x 4x 8x 16x 
    
Sound card:
    Audio device: Realtek ALC892
    
Memory:
    RAM:  7946 Mb
    
Miscellaneous:
    UI Language:  English
    LANG:  en_US.UTF-8
    Microphone:  Not set
    Total Hard Disk Space Available:  112674 Mb
    Largest Free Hard Disk Block:  44820 Mb
    
Installed software:
    
Recent Failure Reports:
    Mon Oct 28 20:13:34 2013 GMT: file ''/tmp/dumps/crash_20131028154840_1.dmp'', upload no: ''Failure when receiving data from the peer''

[Tele42]

Please test with nvidia-driver 331.67, the release notes include "Fixed a bug that could lead to crashes when running Left 4 Dead 2 with threaded optimizations enabled."

[jwcalla]

331.67 didn't fix my issue... but I've found that 331.67 and the previous 331 driver are extraordinarily problematic (for me) in L4D2. The input gets totally screwed up (looped mouse button events or extremely delayed / non-existent mouse button events).

I've never seen anything like it.

[kisak-valve]

Closing as general video driver quality from 2014 is no longer relevant. If anyone is still experiencing this issue with an up to date system, please open a new issue report.