Let me preface this by saying: this is probably a bad motherboard or somesuch, based on the troubleshooting I've done, but I'd just love a sanity check of where I'm at with this...

Alright, so I have a pair of basically identical machines with AMD FX-9590s, we'll call them #1 and #2. The only material differences are:

#1 only has 8GB of RAM, #1 has a GeForce card, #1 has a somewhat inferior CPU cooler.

I've been playing around with Handbrake and other video encoding/editing apps recently and found #1 to be a decent performer for this - it has 8 cores, and the GeForce card also supports NVenc which runs pretty fast - sure it isn't my 32 core workstation, but that's probably an unfair comparison. #1 will sit and run more or less as much as I want video editing/encoding/transcoding/rendering/whatever - it will sit at 100% CPU load, CPU in the high 60* C range, and be fine.

#2 will run at a few degrees lower CPU temperature, and will consistently (As in you can bet on it) hardlock on the desktop 100% unresponsive where the only fix is to kill AC power.

I have tried:
- Swapping PSU
- Swapping RAM (have tried both the 'fancy' RAM it originally had when setup as a gaming box, and some plain-jane Dell-branded DDR3-1333 RAM that's generally very stable in anything)
- Swapping hard disk(s)
- Re-installing and changing distros (so it's tried both Slackware 14.2 x64 and Xubuntu 20.04 LTS)
- Swapping graphics cards (it has a Radeon in it right now)
- Changing case layout to improve airflow/cooling thinking it was a heat issue for the CPU
- Playing with BIOS settings to set more aggressive fan profile, laxer RAM timings, etc (trying to improve stability)

#2 will run CPU stress tests on the bench (100% load x8 threads) just fine, will load up and run games (I've tried Half-Life 2 and Portal) just fine, will decode media just fine, handles web browsing just fine, etc.

At this point I'm thinking it's probably just this motherboard's time - this machine previously ran as a gaming box for a few years, and who knows what before that (I bought the board second hand), but given that it only hardlocks like this when touching video it makes me a bit curious. As it is, I can't trust a machine that will randomly hardlock for anything work related, so while it seems to do 'other tasks' okay, since I can't prove its an issue with a software install vs an issue with hardware, I have to assume the machine is at fault (especially given all the above trouble-shooting).

Does anyone have any other thoughts on this perhaps? Anything I might be missing? I've checked every setting/library/configuration I can against the two - and as far as I can see #2 is 1:1 with #1 (and my big workstation, which is different enough I figure it isn't as fair of a comparison, but nonetheless the settings in userland all look to be the same) by this point (they're all running Xubuntu 20.04 right now, again mostly to try and troubleshoot #2 here).

And finally, the one troubleshooting thing I really have no interest in doing: pulling the CPUs out of #1 and #2 and swapping them. The heatsinks on these chips are very big and very obnoxious to install/remove, and I also don't want to put a known working chip (from a known working machine) into a potentially bad motherboard (and end up with, potentially, two bad CPUs).

If it helps you any. I have used ffmpeg for encoding for years. And let processors run at near 100% on all cores for 2 hours. I've never had encoding a video lock a machine up.

You may try replacing the thermal paste on the cpu heatsink on that machine.

Could that be a leaky capacitor on that board that changes when it gets hot? I don't know how you could test that. You can't check a cap while it is in circuit and those motherboards are multi layer. If you use a solder sucker on them, you could separate a layer.

Looks like you have covered the bases. 60C(140F) isn't too hot for anything. That's just what my quad core i5's do while ecoding.

Is that somehow displaying the video while it is encoding? Lock ups with later kernels and Intel graphics have been a problem this year.
I've not heard of locking with ffmpeg. I haven't used handbrake.

That's what handbrake is now using.
I would not do that either.

How about voltages? You have lm_sensors installed? You could check them against what they should be.
https://hwmon.wiki.kernel.org/lm_sensors

Great ideas - I'll respond in 'multi-quote'

And this is how my experience with encoding has been for a long time - I even remember letting my Q6600 run overnight like that and it was 'just fine' it just took forever. My other two machines are also in this 'things are just fine' - the big workstation the CPUs never even break 50* C!

Should've mentioned: did that when I re-cased it. Basically I 'rebuilt it' just to make sure it wasn't a loose power connector, bad thermal paste, bad mount, etc etc.

See it's this kind of thing that I'm afraid of, and this board has about a million caps/devices on it to boot. Looking in it (with the side panel off) with a flashlight I don't see any bulging caps or burned traces though, FWIW (but like you say, these boards are multi-layer and the damage can be 'hiding within').

And just for my own satisfaction, I literally watched #1 run at the same settings + same temps with no issues...

No it isn't displaying video - I do use the GTK frontend (because I'm lazy and like GUI) but there's no 'preview' while it renders/encodes. Note that this machine has, come to think of it, survived multi-hour long compiling jobs at 80-100% load too...

I don't have Intel GPU in this system so I'm not sure that'd be a problem.



Yes and yes they're reporting sane/normal values - it isn't reporting like 2V Vcore or something absolutely ridiculous.

Update: Now it's getting spooky. I pulled the machine out of the case it was in, put it on my open-air test bench with a (brand)new PSU I have for another build that isn't installed yet, booted it up and it ran the whole job (which has hardlocked it in the past) with no complaints. The CPU also registered 12* C cooler the whole time, and I got to thinking as well that this board has no temperature sensor for the chipset or VRM region (at least that reports to the BIOS or is readable in the OS). I re-enabled the onboard debug LEDs and I'm trying the same job a second time just to see what it does - I had the soundcard out on the 'first run' (because I am lazy.) so hopefully that isn't the 'variable X' here.

Update 2: Two in a row with no problems.

Update 3: Marking [SOLVED] as I've tested the machine through another run or two and had no issues - I'm guessing it was either the bad wiring or insuffucient airflow in the case it was using (I've since re-cased in something with probably too many fans, but it runs quieter and cooler!).