I see the exact same behavior in my issue #1306. Does this crash the Steam client for you as well?
Nvidia GTX 650 Ti running on 313.18.
Great, so 313 drivers are affected too. I think our issues are caused by the nvidia driver or less likely xorg/kernel.
It could be in some OGL 3 stuff, I want to check what game settings exactly causing these problems.
It should be reported to some nVidia dev.
I think we are having different problems then, even though the symptoms look somewhat similar. I'm not having these issues in TF2 and so far it has always involved my Steam client crashing.
OK, I have to update my issue description, since I found lowering settings does not resolve my problem ...
So it could be different bug than #140.
I found when game is stucked even pressing num lock key does not change state of numlock at all - this mean whole kernel is locked. If I set settings to low bug occurance is less frequent.
NOTE: If I press numlock in this locked state, I am able to reproduce #811
OK, it seems I found temporary solution: export __GL_THREADED_OPTIMIZATIONS=0
in hl2.sh script
Now my settings were totally maxed and no issue.
Definitively its driver issue (probably 310 and newer).
OK, today I checked export __GL_THREADED_OPTIMIZATIONS=0 is sometime rewritten to 1.
If I start the game with 0, I can't see any lower FPS, so if this variable is not hardcoded to be still 1, then this optimization does not help at all for me.
But I am afraid __GL_THREADED_OPTIMIZATIONS=0 does not fixed issue completly, it takes only much longer time to freeze (10minutes?).
Its very interesting ... I hope some nvidia dev can fix it.
I emailed our nVidia contacts about this. They haven't seen this issue with their drivers, but they want to collect some data.
Here's what the dev I emailed would like you to do (if you're comfortable with this):
top in thethread apply all bt + info shared?I'm personally running an nVidia GTX 580 and don't see this problem - otherwise I'd gather this data myself.
@bdeen-valve Thank you very much, I agree with everything what could help.
I started CS:Source in window to see updates of top command, screenshot: http://postimage.org/image/8kukf7xzd/
Then I played for around 30 minutes and my pc was only stucked for 2 times with threaded optimization enabled now.
When computer was stucked, my screen was "locked", nothing in my desktop is responding in this time, that mean I can't do alt+tab, change numlock/capslock state and command top was'nt updated on my screen until screen was "unlocked" (after cca 10 seconds). While computer was freezed, I was connected to session with ssh client from other machine and it was responsive, hl2_linux used tiny percent of CPU in this time. Response was probably with little bit delay, but I am sure desktop screen was locked while ssh session not in same time.
If you need I can try to get some gdb info, but in this case it is probably useless.
UPDATE: During hung, Xorg cpu usage was ~104%.
NOTE: I measuring also my GPU temperature and while I am playing there is around 80°C, normally its 55° in desktop. But when crash occured it was 78°C last time.
I am trying to reproduce this issue with TF2, but no luck. I guess there is different engine config, while there I am affected with #140 and #209, in CS:Source it never happened right after map load.
Main difference I see CS:Source is much faster during loading and so does not use same resources as TF2, but in CS:Source something cumulates during gameplay and then its flushed?
Got it too, except it freezes my PC completely. Switching between windows and between desktop and virtual terminals does not work. So it looks like a kernel panic but without any messages. Happens within 1-2 minutes. Nouveau driver works without this issue.
Intel Core 2 Duo 3.0GHz
RAM 4GB
Nvidia GeForce 9600GT 512MB
ArchLinux 64bit
driver version is 313.18
@lxndr: When your PC is freezed are keyboard leds blinking? If so, it is kernel panic, if not, then it is locked screen.
Today I found when my PC is freezed, Xorg is not always taking 100% cpu, but it is rather random. Sometime it is hl2_linux. I think it is just process, which was actually in kernel scheduler?
What's more strange, I can't see any warning or error in dmesg, Xorg log, etc.
All things I did during hung (ALT+TAB, change of num lock state, ..) are done later, respectively after it.
UPDATE: I found mostly there are processes migration/1, migration/2 and migration/3 with around 50% CPU usage.
I don't think they were blinking, but it was not just a screen lock for sure because I couldn't connect to SSH. Forgot to say it doesn't matter if the game is full-screen or windowed. Also last CPU usage was pretty adequate - about 75%. And I could not find anything useful in system logs.
Process of this hung is little bit complicated - It seems for cca first second system is completly freezed (ssh session too). Then ssh is responsive, everything else not, yet. For next few seconds there is Xorg with CPU usage 100% and hl2_linux also with 100% then there are migration processes with around 50% usage and then screen and input events are finally unlocked. (CTRL+ALT+F1 does not work as well during this issue). But it seems some computers are not able to recover from this.
Does not matter I play offline with bots or online.
So high CPU usage of migration process mean there is a problem with scheduling processes between CPUs in kernel.
Could you check if steam is running after hangs? I got this issue ONLY when steam segfaults.
I am sure steam never crashed in my case because of this issue.
Surprisingly I found something is wrong with my graphics card. Today I cleaned GPU Fan what lowered temperature by 10-20°C and my problem is not happening again for now. I don't understand it so much, since TF2 is working without this issue with same temperature. My GPU should handle constantly at least 90°C and it was maximally 80°C. During playing GPU temperature does not exceed 60°C.
If it was caused by GPU overheating accept my apology please.
I'll give a report if it happen again even with cooler GPU.
Should I close this issue? I dont know, because there are more affected ppl with similar symptoms.
All affected ppl should try to clean a fans :)
Would encourage other people that posted on this thread to open new bugs for their behaviour, there’s a long list of posts that make finding true issues harder, thanks!
__GL_THREADED_OPTIMIZATIONS=0x2 2013-02
Hello
Thank you very much for this port, it works nearly perfectly now.
I found only one issue during online gameplay.
All is working fine, until my screen totally freezes for around 10 seconds and last sound sequence starts in loop.
After this, game continue as previously without any issue and 200-300FPS.
With some graphics debugger it could be fine to see what is happening with GPU.
This issue affects me rather randomly, but quite often (1-3 minutes interval), but it was also two times in same place with de_dust2 nearly before bomb plant A.
If I lowered my video settings (shadow and shaders to low, antialiasing off, color correction off) it does not fix issue as in bug #140 with TF2. Only less occurance with low settings.
Is someone able to reproduce this with 313 nvidia driver? I am on 310 and Ubuntu with Gnome 3 Classic, no effects.
My system info: http://pastebin.com/6RDuK4bm