since a few months I encounter the following:
After starting up my desktop with Tumbleweed-KDE I use it on and off during the day. After about 6 hours, suddenly the temperature of the PCI adapter rises and the system is almost unresponsive. (almost: about once every 3 minutes a new value becomes visible in sensors, all the time well above acceptable values)
The command in the console “watch sensors” then shows a temperature of:
composite: 72 C
Sensor 1: 60 C
Sensor 2: 90 C.
Testing promptly after startup with 10 x writing and deleting a video (4 GB) raises the temperature to about 85 C, the system will not hang.
Since looking through Internet I found that KDE is resource-heavy, so I tried a lighter GUI: XFCE.
The test with 10 x writing and deleting gave the same temperature effect.
But even while using the computer on-and-off during 24 hours, no hang occurred.
I would really like to know what happens in KDE that does not happen in XFCE, if only a pointer to a package that causes sudden overheating after about 6 hours.
How clean is the inside of your system? Are the fans and heatsinks, including those inside the PSU, overdue for cleaning? Is there an input air filter that’s never been cleaned? What are the PC specs (inxi -Fz)?
KDE isn’t resource heavy This is a widespread misconception which is is readily demonstrated for my desktop by running “systemd-cgtop -1 --cpu=time user.slice/user-1000.slice/user@1000.service”:
Of course, this when running XFCE
In my earlier reply I put the output of inxi -Fz. I don’t know the correct way: should I just repeat it here or is referral sufficient?
A single fan operates at 488 RPM. Power uptake is 21 Watts when idle. Temperatures stay low when running a youtube video. Provide the output of “systemd-cgtop --cpu=time …” as shown above.
I am unsure whether I need to add: The output of the command is only useful when big temperature spikes have occurred since booting.
As when going to the dentist, pain stops on entering the premises! This is what I found:
After a full night running the system with KDE and using it the next day again on-and-off no overheating occurred. Psensors showed a temparature never over 85 C.
I’ll keep using the system as usual and if normal operation continues will not come back to this thread - at this moment it is too early to consider it solved…
OUCH! after restart this morning, I used the system normally (for me normal = with frequent interruptions) for about 5 hours. Then suddenly, for no visible reasons, the temperature very quickly rose to unacceptable levels resulting in a hang - high temperature visible on Sensor 2 and Composite. I was using Kpat when that happened.
Every few minutes there was a bit of action in the graph of Psensor, whilst the temparature remained high and CPU usage was not extremely high, nor the temperature of Core 0.
I had to leave, so I left the system as it was for about 4.5 hours and on return found still a hanging system with too high temperatures and not excessive CPU-usage (from the scale of the graph - the actual values were off screen) about 40%.
I disconnected from mains (any other method to stop the system failed) and restarted. During the time to restart the computer the temperatures of core 0 had fallen to 67 C, core 1, 2, 3 had fallen to 60, 66, 65 and the temperature from sensor 2 to 55, composite to 50C.
Thereafter I did zypper dup, which resulted in sensor 2 heating to 79 C, CPU usage to 65%, Core 0 to 59C. After the zypper command finished all values dropped to usual levels.
I have no idea what to do further to diagnose the cause of the problem, especially because of the hours long using the system before the hang occurs due to high temperature.
Going away from KDE to XFCE would be possible, of course, but is very unsatisfactory.
Das ist mir ziemlich schnell auf gefallen. Darauf hin habe ich zweimal systemd-cgtop empfohlen. Du hast es zweimal ignoriert. Ich empfehle es zum dritten mal. Danach ist Schluss!
I humbly apologise! Lack of knowledge/understanding hampered me in providing the requested information.
I now believe I understood, here is the output I just now generated:
At this very moment the system is running KDE smoothly.
Psensor shows a very spiky graph with sensor 2 between 32 C and 48C, Core 0 between 30 C and 33 C. Each value remains the same for about 4 seconds max, then either goes up or down, hence the spiky graph.
Accessing Google Maps for a driving time gave a spike for sensor 2 to 52C, which, when doing nothing, slowly (20 seconds) dropped to 38 C; running zypper dup and updating 2 packages raised sensor 2 within 4 seconds to 61 C. After completion it took again about 20 seconds to return to a temperature below 40 C.
To me that suggests that Internet access may be involved, but I have not been able to intentionally overheat-and-hang using the Internet - even an zypper update with more than 500 items did not result in overheating (it remained at about 75 C max). In fact no intentional action resulted in overheating…
CPU times shown by systemd-cgtop are accumulated values since boot. Background slice is fine. Your session slice is quite bloated compared to the app slice. You may compare this to host erlangen:
To further investigate I suggest creating a new user and quitting from the current graphical login. Then login the new user to KDE and run the session until some 1h cpu time has accumulated. Post the new output.
Just after I posted this reply, I had to go to a shop. This took me about half an hour, during which my computer was unattended (nobody home) and quietly doing whatever an idle computer is doing.
Upon return I saw the graph of Psensor with the usual spikes and a significant rise in temperature of sensor2 and composite. The temperature was 78 C. Inspection of the graph showed that the rise started 1 minute before my return and took about 15 seconds to go from 48 to 78 C. The system was slow, but did not hang,so I was able to again run the systemd-cgtop command with the following output:
There is zero difference between two CPU stats which implies that whatever hogged CPU during this period was running outside of cgroup you have chosen.
For today my available time is up - other requests for my time are presented- , comparison of your output to mine I cannot do today anymore, will continue tomorrow.
Nope. You said “Tumbleweed overheats in KDE but not in XFCE”. Thus I suggested to watch [noparse]user@1000.service[/noparse]. However nothing happened in this control group. You need to search all groups for activities: