This is making no sense at all. None. Zero.

I decided to overcook the CPU, just to check it still issues the shutdown -r. But I can't overheat it!

Just sitting on one agetty tty, the cpu is up at 52º. I thought I knew the Bios put the fan into overdrive at 70º. I was wrong. The fan only ticks over below 80º. It revs up at 82º, and goes into overdrive at 86º So I'm getting cooled off on a warmer day, and once I get up to 86º, it drives the temperature back down to 81º Despite a higher load,I couldn't get up to yesterday's reading of 88º

I ran mksquashfs on a specially arranged 8G iso, ran videos and watched them. Top showed ≅390% , but it wouldn't overheat. I tried a kernel build too. Running anything besides mksquashfs actually lowered the perceived load, because make (using two cores) subtracted from the % on four threads I even tried blocking the cooling vents, but it managed anyhow, presumably by dropping the freqency. Then I thought "If the sensors are lying to me, I could be actually destroying this thing." So I gave up. Trying to load the GPU does nothing - I'll have to get a video 3d test of some sort.

Apart from saying that my laptop is on the ropes, I can't draw any conclusions about what is going on except that I have been made a fool of by an inanimate machine. That aside, I haven't an iota why I get a hard poweroff from Zoom. It seems to be me, and only me, as my friends on (Spit!) M$ windows, MacOS, iOS & Android all stay put.

Phoronix is doing these hardware tests, I'm sure they have a Linux utility for GPU testing, and maybe even answers to your laptop problem.

But, since you are unable to reproduce the problem otherwise, back to the software:
what distro, how did you install Zoom, and a list of installed files.
Try Zoom with a new profile/config.
Etc.

This is literally where I got to with the desktop box I used for Zoom (against my will) for the last year - except it ran nowhere near this hot, had much more power available to it, etc (that is, I could control out all of the things you're limited by on a laptop), and it would happily sit and run 100% CPU load 24x7, light the GPU up, etc and not crash, but then along comes Zoom and 'it just crashes.' I gave up and blamed Zoom as junk-ware and accepted 'it will just crash' as the most technical outcome. As far as 'why it worked before but not now' - because Zoom forces updates both server and client side constantly with almost no documentation would be my guess - 'they changed something' and now 'it doesn't work' and that's as much as you'll ever see about it. I'm not sure on the Linux side but I know on macOS and Windows they were/are shipping rootkit-esque backdoors ('for user experience convenience') which may be futzing with power or stability, Apple pushed patches to break it (and restore system sanity) on macOS, not sure the state of things beyond that though.

On the temps: that isn't surprising - newer CPUs will generally not let themselves overheat, so they will throttle pretty aggressively as they come right up to the line, but when you get into doing things like 'blocking output vents' and so forth you are risking damaging the box, as you point out. The rest of it isn't surprising though - you'd have to watch clockspeed and core voltage as you did it to see exactly what is going on, but it should basically do 'anything it can' to keep itself under 90*C or whatever the max point is, even if that means dropping to sub-1GHz clocks. Intel GPUs tend not to make very much heat overall, so loading that up shouldn't make a huge difference (and it probably uses different parts of the chip for video and 3D to be more efficient).

Also unrelated but worth pointing out: Waterfox got bought out by an advertising/surveillance company a few months back, and should be treated with the utmost skepticism (if not just outright abandoned for Firefox) as their 'principles' are nonexistent.

Got it.

The cooling fan's speedup temperature is drifting around the place, so thermal cutouts can probably be activated. I'm observing it, but don't ask me why. Today, I'm getting into top fan speed in the low eighties, so I can't overheat the thing. CPU load is hard to judge, because load builds up as video is being streamed. They may be doing some software rendering, which would account for the high cpu usage. I'm seeing 175% - 201% of cpu to Zoom alone.

On reflection, it was pretty ridiculous (yesterday) to have the fan hitting top speed at 86º and the cpu going critical at 90º. But my box is in a North facing room which stays pretty cool all year round. My cpu is 35W, whereas most of these model have 55W cpus.

Sorry guys - replying to posts

@obobskivich: I will accept crashes - fine. Powering off with the fan still spinning down is a hardware fault, to my way of thinking. My laptop has power (mains & battery) but the motherboard would have to be instructed not to use the power it has, and I don't think that can be done. It can be cut at the step-down stage (i.e. 19V ==> 5V/3.3V), which I suspected earlier. On standard PC power supplies, there's a "power good" line, and if that goes, you powered off with the fans spinning down. It was a crude attempt to save the motherboard. Whether the laptop has one, I don't know. In my case, the 19V is present even without mains. So, reliability is gone.

Is your box old? Can you swap out your power supply?

@Ondoho: Yes, phoronix do their test suite. It's a bitch to install and even worse to use. I'm on Slackware64-Current iso of April 20th. Slackware-15.0 has gone beta, but the beta was the iso 8 days previous. I install the 'other linux' tar.xz, which wants it's own directory in /opt. I updated just now
from version 5.5.7938.0228 to 5.7.25958.0621, but I haven't tried it yet, beyond a 'smoke test.' The smoke test was a thing in electronics; If you switched on and powered up your new prototype, sat there for 30 seconds; if parts didn't start smoking, you had passed the smoke test. Whether the prototype worked even vaguely as you intended was another step altogether.

As Zoom distribute the tar, I don't think the file list is relevant. I have no fancy profiling in the config. I'd like to set up a background but usually spend time cringing any time one of my friends tries it. You seem to get awful faults, mainly coloured smears or tears in the space they have just left. I have the distinct feeling the linux version was contracted out to some newbie who is keeping quiet and asking very basic questions here

I'm not sure what to tell you - on linux these kinds of 'crash to reboots' are admittedly very rare, but on Windows they can be par-for-the-course with certain types of applications (games especially). No idea on whether or not the laptop has something approximating 'power good' (or any other ATX-style features) internally.

On the box I tested this with: I've swapped every piece of hardware through (or tested what I can independently, as applicable), Zoom will still randomly hardlock/crash-to-shutdown/reboot the machine even when run in-browser. Remove Zoom, remove the problem. My guess is that its trying to read/write somewhere it shouldn't be (I don't mean 'security' I mean 'is garbage' or 'is insane'), and creating instability.

The box itself is 'old' but not so awful it can't run this - it has an AMD A10-5800K, and I've tried with an nVidia dGPU as well (seems to make it 'more stable' but this is still Windows 98-level user experience IME, so it stayed (yes past-tense, and I'm happy to be free of this curse) on its own dedicated box that does nothing else whatsoever specifically because of the crashiness).


Oh, I thought it was just me. At least I'm not alone.


You mean the entire application? I still don't understand how this nobody company with zero market share became, overnight, the 'world standard' for video communication against multiple established players in a saturated market, and they didn't offer anything that was really novel or revolutionary versus the mature platforms. I feel like those questions are probably above my paygrade though. Given the amount of complaints and issues they caused on the Mac and Windows side as well, I think the whole thing is just a ball of mud, but hey, I'm sure the C-suite got huge bonuses and think of all the shareholder value!

Laptops in the main don't make real attempts to conform where they don't have to. In the 1980s pc compatibility was a big thing, but even a decade later it was taken for granted.

In March 2020, our guys never considered google hangouts (or whatever their app is). Nor did they consider Jitsi. They're catering for a low tech worldwide audience and came up with Apple face time or zoom. Zoom was much worse then than now. A free M$ download would have been a big consideration.

I updated to zoom-5.7.25958.0621 yesterday. There's a version.txt in zoom which gets you your version. Who knows? But you need an account with them to report bugs.

Yes, I'm sure their ship has come in, as the timing was excellent. Here the delivery sucks as they are on Amazon servers which can be as crowded as rush hour traffic in the evenings.

Well, I finally got on to a big Zoom meeting lasting 2 hours.

My temperature sat in the mid 80sºC, within a degree or two of where the fans start blowing hard, and it didn't shut down in 2 hours. mid 80sºC is perilously close to the 90ºC shutdown temperature. So instead of (from memory) the original 4 fan thresholds being set at

1. 30ºC
2. 40ºC
3. 60ºC
4. 70ºC

my box is now seeming to control temperature at these spots.

1. ??ºC
2. ??ºC
3. 82ºC
4. 86ºC

I'm not seeming to see the lower two thresholds because the CPU is always above them, and the fan is so quiet it's difficult to hear them.

The 82º & 86º are highly suspect. Nobody would design cooling so near to the critical temperature. You would run there if you wanted to keep the cpu AT critical. So that's my hardware problem. The output from 'sensors' shows Furthermore, there's now hysteresis in what was a straight proportional control. Previously, fans blew hardest at 70º and above. Now fans blow hardest at 86º, and stay blowing hard until I get own to 81º

Looking at the "high = +72.0°C," 70ºC would be a logical place to send the fan into overdrive, which is where I remember it.

So my 9 year old box is ropey, the fault is pointing at the CPU area, or concievably BIOS, and I'm not going there. I'm not hardware hunting yet, but that will happen later, as there's an interesting year ahead which may see cash being splurged in a number of areas. I'll let some of that pan out before buying a box.

I had a similar issue with a laptop and it turned out to be heat related.
The air flow path inside the machine had become partially blocked with lint and at low activity levels the machine would run for extended periods.
When I bumped up to >50% CPU activity the system would run fine for some small amount of time then suddenly be off.

I figured out it was inadequate cooling. At higher loads the cpu would overheat so rapidly that it was not even able to be seen by sensors so was difficult to actually track down. The cpu was doing the safety shutdown and there was never even a hint in the logs due to the suddenness of the shutdown.

I have seen in the bios of newer systems that there is often a way to change the fan settings to numbers that suit your usage. Maybe that is possible for you as well.

I think that running at 80C+ for extended time is dangerous

Actually...


Most modern systems (especially embedded/mobile) are designed to run right at the line like that, and use clock throttling on the CPU for fast adjustments to keep it right at the line, so that they can advertise as being 'very quiet' systems. This is also true for a lot of graphics cards' stock fan curves. I agree with you that it's not sane, but it is very typical.

Can you take control of the fans on the machine from software and help it out some? I know there is a cli based fan control program that started out for Thinkpads and has spread to controlling most any fan, but the exact name escapes me at the moment ('thinkfan' or something like that). If you're using an AMD APU it may also be controllable via Radeon Profile, which has a GUI.

Can you open this thing and try cleaning dust out and repasting the CPU/SoC? I do realize that's a 'huge ask' depending on the machine.

Practically, no. I've only one arm working; it's a 17.3" screen and I can't lug it about safely; My rather neat toolboxes have degenerated into chaos, as my son now does the work; the screws in the thing really need replacement; I could manage, but he won't. I could hand it in, but I'm not likely to.

Maybe that temperature change came with a BIOS update, which I think I did. Mine is an anything-but-modern i3 based system. The CPU has very primitive temperature control and temperature sensors remote from the wafer, A semiconductor junction makes a good temperature sensor which can be put in the hottest spot, but mine seem mechanical. So it couldn't perform reliably at those temperatures.

In fact, that last sentence sounds like the problem is right there.

Yes a BIOS update would (could) change the fan profiles. Can you control the fans yourself? I found the project page for the software I was thinking of: https://github.com/vmatare/thinkfan

It cut out again, then yesterday after 20 minutes. I restarted and i ran for over an hour. So I guess I can eliminate temperatures and narrow it to software. It's also spitting up windows the totally wrong size and when I'm resizing them in the early stages. Zoom points out (in less clear terms) that excessive CPU usage is degrading my user experience. It sounds like it's doing software rendering?

I found a small thermal vent underneath the pc, so I have it propped up on a few cdr blanks, and it seems to be running a few degrees cooler.

From your description here I think it very certainly can be temperatures. Propping it up to allow better airflow under the laptop and longer operating time gives me that impression.

A laptop cooler that raises the body up from the surface usually has fans and air vents under the main laptop body to improve cooling. They are relatively inexpensive and may make a world of difference if used. The one I have is usb powered so I can carry it anywhere the laptop can be used.

To see what the temps are actually doing you might install lm-sensors, then the sensors-detect command will configure it for what sensors the machine has and the sensors command can give you a lot of detail on actual temps, fan speeds, etc. Not all machines support the same sensors so what you see will be hardware dependent.

You can add gkrellm as well if you want a real-time graphical display of the sensors output.