[SOLVED] System freezing. Not sure about cause. Possibly video card?

magister94 · 2017-07-07 03:52:20

I've got a fresh arch install. The second time I've installed arch and I think it wen't fairly smoothly.

I tried running the game Factorio and at a seemingly random time after starting the whole system would freeze. sometimes within 1 minute, sometimes 3.
I ran Metro 2033 Redux to see if it was gaming in general and it froze before I could get into the actual gameplay.

So I switched graphics drivers from xf86-video-amdgpu to xf86-video-ati.
I thought the issue was fixed when I was able to successfully run Factorio and Metro at the same time for over 15 minutes. The next day, I restarted chrome and the same freeze happened. Then later, it froze from factorio after a few minutes.
And in Metro again, so It appears the issue is still present and not fixed.

I'm very new to debugging this sort of thing.
I've enabled sysrq keys and they don't seem to work during this freeze, so I assume that means the kernel is frozen.
On reboot, I run

journalctl -b -1

to see if anything stands out, but there's nothing obviously wrong and the last lines are never the same and look benign.

I'm not sure how to continue investigating this problem.
Any help/advice would be much appreciated.

Here's some system information.

Linux 4.11.9-1-ARCH #1 SMP PREEMPT Wed Jul 5 18:23:08 CEST 2017 x86_64 GNU/Linux
XFX Pine Grenada XT [Radeon R9 390X]
Intel(R) Core(TM) i7-5930K CPU @ 3.50GHz

I'm using Gnome, wayland, mesa drivers with xf86-video-ati.

EDIT: turned out it was the CPU overclock. Seemed stable in windows, wasn't in Arch.

Last edited by magister94 (2017-07-09 06:13:17)

magister94 · 2017-07-07 04:27:54

Just tried the this instability fix here https://wiki.archlinux.org/index.php/AT … nstability and it still froze.

HiImTye · 2017-07-07 04:35:05

can you switch TTYs or SSH into it?

magister94 · 2017-07-07 04:51:04

HiImTye wrote:

can you switch TTYs or SSH into it?

Can't switch TTYs. Haven't tried SSH, but SysRQ commands also don't work.

seth · 2017-07-07 06:22:55

Mindstorm:
- Did you really enable sysrq? (ie. "i know what I do and didn't just follow a youtube tutorial from some ubuntu hipster" ;-) - It doesn't even reboot?
- Temperature issue?
- Self-assembled box? Does the GPU come with a dedicated power supply (and is it supplied?)
- Does it happen on X11 as well (and maybe try openbox rather than gnome)?
- Does sound keep running?
- Flashing kbd LEDs?
- Also run "dmesg -w" in sight when waiting for this to occur (ie. along chromium, not a game - so you can actually see the dmesg window in action)

magister94 · 2017-07-07 07:02:05

seth wrote:

- Did you really enable sysrq? (ie. "i know what I do and didn't just follow a youtube tutorial from some ubuntu hipster" ;-) - It doesn't even reboot?

I followed the wiki and tested the key commands while the system was function properly to make sure I could actually get it to restart when it's not frozen. Also have sysrq enabled on boot and verified that is working.

seth wrote:

- Temperature issue

I don't think so. Factorio is not a very demanding game. and sometimes it will freeze in less than a minute, or before I even get out of the menu and into the game. But earlier, I spent 20 min in Metro before it froze, a massively demanding game.
Also, I don't have this issue in windows.

seth wrote:

- Self-assembled box? Does the GPU come with a dedicated power supply (and is it supplied?)

Self-assembled. 1000watt psu. and as previously mentioned, I don't have this issue in windows, which suggests to me a software issue. (or a hardware issue that doesn't surface in windows?)

seth wrote:

- Does it happen on X11 as well (and maybe try openbox rather than gnome)?

I haven't tried. I'm not entirely sure how to go about switching from wayland to xorg. Haven't put the effort in to try. I was hoping to not have to rely on xorg, and if that means finding issues in wayland/gnome/whatever then that's what I'd like to pursue.
But, I suppose you mean for me to try changing things out to narrow down the problem area.
I'll try to figure out how to switch gnome to xorg instead of wayland. If that doesn't fix it, then I'll check out openbox. Will report back results.

seth wrote:

- Does sound keep running?

No. and no loop of the last bit of audio either.

seth wrote:

- Flashing kbd LEDs?

I'm not sure what you're talking about. Googling suggests some sort of heartbeat LEDs with keyboard backlights? But not sure I've found what you're talking about. Could you point me to a resource to learn more?

seth wrote:

- Also run "dmesg -w" in sight when waiting for this to occur (ie. along chromium, not a game - so you can actually see the dmesg window in action)

Will do. I forgot to mention, I have 3 monitors. So I'll do this with a game and a terminal on one of the other monitors immediately after posting this and return with results.

seth · 2017-07-07 07:16:23

The kernel starts to flash/blink the keyboard status LEDs (num lock, caps lock and scroll lock) on a kernel panic.
The problem *could* be that the 3D unit (for whatever reason) only goes full steam on linux while being more adaptive under windows - so I'd not rule out insufficient power supply (not the 1000W, but a dedicated power line to the GPU next to PCI power) or temperature issue just because it works on other OS.

magister94 · 2017-07-07 08:29:02

dmesg -w results.
The only thing that happened while the game was playing was this message at launch.

[  101.385988] warning: process `metro' used the deprecated sysctl system call with 10.1.

nothing happened when it froze.

I noticed that the audio loops when I wear headphones plugged into the front 3.5mm audio out, but not when using the digital optical out to speakers.

To see if temperature was the issue, I installed lm_sensors and ran `watch sensors` while playing Metro.
Throughout the gameplay it was around 82 - 84c and was 84c when it froze.
From what I understand, anything below 85c is normal for this card.

The kernel starts to flash/blink the keyboard status LEDs (num lock, caps lock and scroll lock) on a kernel panic.

if this is something that happens automatically, then no. It's not happening to me.

so I'd not rule out insufficient power supply (not the 1000W, but a dedicated power line to the GPU next to PCI power)

Not sure what you're referring to. the 8 and 6 pin connectors to the GPU? or power from the mobo?

I just tested Xorg by logging in while selecting "GNOME on Xorg".
It froze before I got to the games main menu, at a temp of 52c for the GPU.

Do you still think I should still try openbox? Do I need to remove gnome first? I see some stuff about "Openbox inside the GNOME desktop environment"

seth · 2017-07-07 12:07:22

8/6pin connectors, yes. If your GPU relies on them, it starts acting heuristically when underpowered.

This bug suggests DPM issues with the GPU
https://bugs.freedesktop.org/show_bug.cgi?id=91880

Also the GPU might be a red herring - do you have intel-ucode installed and enabled?

magister94 · 2017-07-07 23:38:03

seth wrote:

do you have intel-ucode installed and enabled?

I do not.
https://wiki.archlinux.org/index.php/Microcode
I suppose I should?

V1del · 2017-07-07 23:52:53

Yes. Many strange crashes are due to issues with the microcode especially for Broadwell CPUs like yours.

magister94 · 2017-07-08 00:18:10

installed it. Still freezing just like before.

verified the microcode was actually applied with dmesg | grep microcode and grep microcode /proc/cpuinfo.

[    0.000000] microcode: microcode updated early to revision 0x3a, date = 2017-01-30

Last edited by magister94 (2017-07-08 00:24:31)

magister94 · 2017-07-09 06:10:22

Well this is embarrassing.
I think the problem was my CPU overclock. I completely forgot I had done that. It seemed entirely stable in windows for a long time. Disabled the overclock and now I was able to play Metro for over an hour without issues.
I guess linux takes advantage of the CPU in ways windows doesn't.

EDIT: thanks to everyone who tried to help. I appreciate it and I learned some stuff.

Last edited by magister94 (2017-07-09 06:16:41)

Arch Linux

#1 2017-07-07 03:52:20

[SOLVED] System freezing. Not sure about cause. Possibly video card?

#2 2017-07-07 04:27:54

Re: [SOLVED] System freezing. Not sure about cause. Possibly video card?

#3 2017-07-07 04:35:05

Re: [SOLVED] System freezing. Not sure about cause. Possibly video card?

#4 2017-07-07 04:51:04

Re: [SOLVED] System freezing. Not sure about cause. Possibly video card?

#5 2017-07-07 06:22:55

Re: [SOLVED] System freezing. Not sure about cause. Possibly video card?

#6 2017-07-07 07:02:05

Re: [SOLVED] System freezing. Not sure about cause. Possibly video card?

#7 2017-07-07 07:16:23

Re: [SOLVED] System freezing. Not sure about cause. Possibly video card?

#8 2017-07-07 08:29:02

Re: [SOLVED] System freezing. Not sure about cause. Possibly video card?

#9 2017-07-07 12:07:22

Re: [SOLVED] System freezing. Not sure about cause. Possibly video card?

#10 2017-07-07 23:38:03

Re: [SOLVED] System freezing. Not sure about cause. Possibly video card?

#11 2017-07-07 23:52:53

Re: [SOLVED] System freezing. Not sure about cause. Possibly video card?

#12 2017-07-08 00:18:10

Re: [SOLVED] System freezing. Not sure about cause. Possibly video card?

#13 2017-07-09 06:10:22

Re: [SOLVED] System freezing. Not sure about cause. Possibly video card?

Board footer