[Solved] System freeze even with light browsing

thepanu · 2019-06-12 20:13:17

I am facing a freeze that I am not able to figure out. It can happen even after 15 minutes or less of light browsing. Display will freeze, caps locks won't respond, network drops, can't ssh in. REISUB boots the computer just fine in 9/10 cases, but sometime even that doesn't help and reboot needs pressing the reset button.

In this Gist there is is one complete boot and freeze cycle.

https://gist.github.com/thepanu/3e1fa77 … 626849a095

I had similar issue earlier (https://bbs.archlinux.org/viewtopic.php?id=240387) but that was fixed when I RMA'd my PSU. With the new PSU my system worked great for many months but recently this issue came up. Nothing in the hardware has been changed and everything is less than 1 year old.

I ran memtest and got 17 clean passes in 41 hours at which time memtest had frozen so that timer wasn't running anymore nor was anything else happening other than blinking + in top left. Comp did rebooth when ESC was pressed.

Any and all advice is appreciated.

Last edited by thepanu (2019-10-29 20:08:40)

seth · 2019-06-12 21:05:19

at which time memtest had frozen … blinking + in top left

The the GPU didn't stall (it's in VGA mode anyway) but memtest looped out. Downclock the system, in particular the memory timings (all the way down, the most conservative settings), and see what happens.

thepanu · 2019-06-13 16:02:06

I am running my memory at advertised settings, haven't touched timings etc. But that has been selected with XMP, now I returned it to automatic settings and ran mprime torture test blend for 1,5h without issues. After canceling that and began to type this message my system froze almost immediatly. More testing if I can find a way to force the issue.

seth · 2019-06-13 16:21:45

Temperature?

thepanu · 2019-06-13 19:22:38

CPU temp doesn't go over 60 even under full load (mprime). I also ran stress-ng for an hour aiming for 100% utilization and no issues. After that I played Mordhau for 20min and again a freeze. There doesn't seem to be any correlation between load levels and freeze.

I noticed that there is new bios version available so I'll give that a go but I don't have high hopes for that as this has been stable on the current version.

EDIT: Another crash while browsing (firefox). I had youtube video playing on the background and the sound played a minute or so after everything else froze. This was with the new bios and default memory speed.

Last edited by thepanu (2019-06-13 20:02:27)

euromatlox · 2019-06-13 20:05:25

If replacing PSU has corrected similar problem before, I would try that solution again. And perhaps a slightly more powerful PSU (same Voltage(s), more Amperes) might be good idea.

If it's desktop PSU (metallic box) you can try to look through ventilation holes using flashlight if any leaking capacitors can be seen. Voltage readings might also give a clue, those can be also measured using multimeter to be sure.

BIOS battery (CR2032) is good to check too.

Last edited by euromatlox (2019-06-13 20:18:11)

thepanu · 2019-06-13 20:08:04

That may be the way unfortunately.

seth · 2019-06-13 20:10:50

Mordhau and FF sounds like GPU related.
Does yours have a dedicated power supply (6/8 pin plug)?
Does it reside in the proper PCIe slot (consult the board manual - no, they're usually not all the same)

No idea how to get memtest … wait a second: Is your OP about memtest86+ or about "memtest" mprime?

thepanu · 2019-06-14 17:02:01

Yes, there is dedicated plug. And it is in the only 16x slot the mobo has.

OP was about memtest86+. Since then all tests have been on desktop and on mprime I used torture test blend.

Sounds like GPU to me as well. I finished 16h run of following simultaneously

stress-ng --vm 1 --vm-bytes 98% --vm-method all --verify -t 16h
stress-ng --cpu 0 --cpu-method all -t 16h

Although I don't think it is 'real' load even though htop shows everything at max as temps don't rise nearly as much as with mprime.

No problems during the run when I had only couple terminals and psensor open. After the test I launched FF and got an immediate freeze.

Last edited by thepanu (2019-06-14 17:02:48)

seth · 2019-06-14 18:58:14

Can you switch the GPU (prefereably to a low end one w/ low/er power demands)?

thepanu · 2019-06-14 19:33:34

Not immediately no, as I don't have one at hand, sadly I just gave a way my last extra card. My calculated power need should be around 350W and I have 650W psu.

seth · 2019-06-14 19:38:24

I'm rather worried about the current on the dedicated plug (or since you changed the PSU the slot! maybe also the PCIe connection/seating) than about the PSU output.

Assuming the memtest freeze was just a fluke and if supporting your GPU, you could also try the 390xx drivers.

euromatlox · 2019-06-14 22:26:16

I assume it's a single rail (12V) PSU, since if it's multi rail (12V1,12V2,12V3,...) PSU, there's always a possibility to have too much load on a rail. There was a time period, when most PSUs where multi rail ones.

Last edited by euromatlox (2019-06-14 22:29:31)

thepanu · 2019-06-15 16:29:34

FWIW it is this psu https://www.silentiumpc.com/en/product/ … gold-650w/

I'll try reseating everything next.

Thanks for the support!

thepanu · 2019-06-16 18:12:00

Quick update. A totally new kind of crash yesterday, before I got to reseating. Again playing Mordhau and suddenly total crash.

Signal to display off (display goes to stand by)
doesn't respond to sysrq boot, which has worked pretty consistently
doesn't shutdown from power button.

I cycle the the switch in the psu and immediately the fans come back on.

I turn psu off and decided to wait a bit longer. After something like 20s there is small click, like a relay or similar and then I turn it back on from psu switch. Fans don't turn. Boot works fine after that.

Last edited by thepanu (2019-06-16 18:12:34)

thepanu · 2019-10-21 17:17:35

After many months of just living with the crashes I come back to this issue. I recently bought a new gpu as most often logs pointed towards gpu issue (GPU has fallen of the bus etc). My old gpu was Nvidia GTX970 which I replaced with AMD RX590.

If there is any change then it is that crashes now come more often. Other than that the biggest change is there is nothing in the logs.

After changing memory, rma'ing and replacing psu, rma'ing cpu and replacing gpu with another brand I still face the same issues. Which leaves mobo. Guess I might as well try that next.

FWIW only kernel parameter at the moment is 'loglevel=8'.

Last edited by thepanu (2019-10-21 17:20:44)

Pse · 2019-10-21 19:03:45

Have you tried moving the GPU to a different PCI-E slot? There may be some electrical issue with that specific slot.

thepanu · 2019-10-22 06:39:15

I did, but the only other slot is x4 and boot ended up in Kernel panic with that.

I did manage to get one crash with something in the error logs, a PCIe bus error, BadTLP.

https://unix.stackexchange.com/question … er-bad-tlp

As per this I tried to boot with pci=nommconf, but no difference in results. Fastest crash I have had is while typing in password to log in.

thepanu · 2019-10-29 20:08:25

Marking as solved. Replacing Mobo stopped crashes. While earlier it would crash as early as in login now I have had the system up and running 4 days straight with no issues whatsoever.

Arch Linux

#1 2019-06-12 20:13:17

[Solved] System freeze even with light browsing

#2 2019-06-12 21:05:19

Re: [Solved] System freeze even with light browsing

#3 2019-06-13 16:02:06

Re: [Solved] System freeze even with light browsing

#4 2019-06-13 16:21:45

Re: [Solved] System freeze even with light browsing

#5 2019-06-13 19:22:38

Re: [Solved] System freeze even with light browsing

#6 2019-06-13 20:05:25

Re: [Solved] System freeze even with light browsing

#7 2019-06-13 20:08:04

Re: [Solved] System freeze even with light browsing

#8 2019-06-13 20:10:50

Re: [Solved] System freeze even with light browsing

#9 2019-06-14 17:02:01

Re: [Solved] System freeze even with light browsing

#10 2019-06-14 18:58:14

Re: [Solved] System freeze even with light browsing

#11 2019-06-14 19:33:34

Re: [Solved] System freeze even with light browsing

#12 2019-06-14 19:38:24

Re: [Solved] System freeze even with light browsing

#13 2019-06-14 22:26:16

Re: [Solved] System freeze even with light browsing

#14 2019-06-15 16:29:34

Re: [Solved] System freeze even with light browsing

#15 2019-06-16 18:12:00

Re: [Solved] System freeze even with light browsing

#16 2019-10-21 17:17:35

Re: [Solved] System freeze even with light browsing

#17 2019-10-21 19:03:45

Re: [Solved] System freeze even with light browsing

#18 2019-10-22 06:39:15

Re: [Solved] System freeze even with light browsing

#19 2019-10-29 20:08:25

Re: [Solved] System freeze even with light browsing

Board footer