Overheating problem with Ubuntu OS

[QIII]

Is this a dual-boot system. Does it still behave normally in Windows if it is?

Have you recently opened the machine and perhaps bumped the cooler?

[Doug S]

You can observe the status of thermald via:

doug@s15:~/idle$ sudo systemctl status thermald.service
[sudo] password for doug:
● thermald.service - Thermal Daemon Service
   Loaded: loaded (/lib/systemd/system/thermald.service; enabled; vendor preset: enabled)
   Active: active (running) since Wed 2018-12-19 23:50:55 PST; 2min 52s ago
 Main PID: 13755 (thermald)
    Tasks: 2
   Memory: 1.5M
      CPU: 23ms
   CGroup: /system.slice/thermald.service
           └─13755 /usr/sbin/thermald --no-daemon --dbus-enable

Dec 19 23:50:55 s15 systemd[1]: Starting Thermal Daemon Service...
Dec 19 23:50:55 s15 systemd[1]: Started Thermal Daemon Service.
Dec 19 23:50:55 s15 thermald[13755]: 13 CPUID levels; family:model:stepping 0x6:2a:7 (6:42:7)
Dec 19 23:50:55 s15 thermald[13755]: Polling mode is enabled: 4
Dec 19 23:50:55 s15 thermald[13755]: XML zone: invalid sensor type []

You can re-start it via:

sudo systemctl restart thermald.service

I re-installed it on my test server and set the trigger temperature to 55 degrees. I fully loaded all of my CPUs with thermald disabled, then enabled it. Turbostat output:

doug@s15:~$ sudo turbostat --Summary --quiet --show Busy%,Bzy_MHz,PkgTmp,PkgWatt --interval 15
Busy%   Bzy_MHz PkgTmp  PkgWatt
100.00  3400    66      57.75
100.00  3400    67      58.23
100.00  3400    69      58.59
100.00  3400    70      58.85
100.00  3400    71      58.97
100.00  3199    65      54.54  <<< Thermald service started.
98.96   1678    58      23.68
71.25   1600    55      17.26
69.45   1600    54      16.82
88.33   1600    54      20.33
100.00  1600    52      22.44
100.00  1909    54      27.60
100.00  2114    55      31.15
100.00  1743    56      24.72
100.00  1862    51      26.85
100.00  2045    56      29.91
100.00  2035    53      29.79
100.00  1869    55      26.99
100.00  1945    53      28.20
100.00  1909    54      27.62
100.00  1770    55      25.23
100.00  1817    54      26.08

EDIT: I am using the intel_pstate CPU frequency scaling driver and the powersave governor. For less extreme temperature control thermald limits CPU via the max percent setting:

doug@s15:~/idle$ cat /sys/devices/system/cpu/intel_pstate/max_perf_pct
60

If the temperature is still too high, even at minimum CPU frequency, then thermald starts to use "kidle_inject", which is why you see my "Busy%" drop below 100%, for a short time, above. Also:

doug@s15:~$ sudo turbostat --quiet --Summary --hide Avg_MHz,SMI,GFXMHz,TSC_MHz,GFXWatt,CorWatt,IRQ,POLL,C1,C1E,C3,C6,CPU%c1,CPU%c3,CPU%c6,CPU%c7,CoreTmp,GFX%rc6,Pkg%pc2,Pkg%pc3,Pkg%pc6 --interval 5
Busy%   Bzy_MHz POLL%   C1%     C1E%    C3%     C6%     PkgTmp  PkgWatt
100.00  3400    0.00    0.00    0.00    0.00    0.00    74      59.69
100.00  3400    0.00    0.00    0.00    0.00    0.00    74      59.74
100.00  3400    0.00    0.00    0.00    0.00    0.00    75      59.69
100.00  3400    0.00    0.00    0.00    0.00    0.00    75      59.69
100.00  3400    0.00    0.00    0.00    0.00    0.00    74      59.68
100.00  3400    0.00    0.00    0.00    0.00    0.00    76      59.74
100.00  3400    0.00    0.00    0.00    0.00    0.00    75      59.72
100.00  3400    0.00    0.00    0.00    0.00    0.00    75      59.74
100.00  3241    0.00    0.00    0.00    0.00    0.00    71      56.11   <<< Thermald enabled
100.00  2358    0.00    0.00    0.00    0.00    0.00    65      36.86   <<< Thermald starts by lowering the maximum CPU frequency
100.00  1712    0.00    0.00    0.00    0.00    0.00    62      24.71
99.88   1600    0.00    0.00    0.00    0.00    0.12    61      22.77    <<<< Processor is still too hot, even at minimum CPU frequency. Thermald starts using kidle_inject
92.67   1600    0.00    0.00    0.00    0.00    7.34    60      21.36    <<<< Kidle_inject forces C6, the deepest and lowest energy idle state (for mine, and your, processor)
80.66   1600    0.00    0.00    0.00    0.00    19.35   58      19.07
61.64   1600    0.00    0.00    0.00    0.00    38.38   56      15.49
50.57   1600    0.00    0.00    0.00    0.00    49.45   56      13.51
49.85   1600    0.00    0.00    0.00    0.00    50.17   55      13.35  <<<< Minimum CPU frequency and 50% idle.
49.85   1600    0.00    0.00    0.00    0.00    50.17   55      13.33
49.85   1600    0.00    0.00    0.00    0.00    50.17   53      13.28
49.85   1600    0.00    0.00    0.00    0.00    50.18   52      13.28
49.85   1600    0.00    0.00    0.00    0.00    50.18   52      13.23
51.60   1600    0.00    0.00    0.00    0.00    48.42   51      13.56  <<<< Temperature undershoots target
57.74   1600    0.00    0.00    0.00    0.00    42.28   51      14.71
64.30   1600    0.00    0.00    0.00    0.00    35.71   51      15.86
71.26   1600    0.00    0.00    0.00    0.00    28.75   51      17.17
77.15   1600    0.00    0.00    0.00    0.00    22.86   52      18.18
83.29   1600    0.00    0.00    0.00    0.00    16.72   51      19.30
89.44   1600    0.00    0.00    0.00    0.00    10.57   52      20.44
96.26   1600    0.00    0.00    0.00    0.00    3.74    52      21.70
100.00  1600    0.00    0.00    0.00    0.00    0.00    52      22.38  <<<< Thermald has now removed kidle_inject
100.00  1600    0.00    0.00    0.00    0.00    0.00    52      22.41
100.00  1600    0.00    0.00    0.00    0.00    0.00    52      22.38
100.00  1668    0.00    0.00    0.00    0.00    0.00    53      23.42  <<<< And even starts to increase the CPU max Frequency
100.00  2071    0.00    0.00    0.00    0.00    0.00    56      30.24
100.00  2050    0.00    0.00    0.00    0.00    0.00    53      29.98
100.00  2220    0.00    0.00    0.00    0.00    0.00    55      33.10
100.00  1832    0.00    0.00    0.00    0.00    0.00    53      26.17
100.00  1600    0.00    0.00    0.00    0.00    0.00    52      22.40
100.00  1788    0.00    0.00    0.00    0.00    0.00    55      25.44  <<<< It is not a great servo system, so oscillates some.
100.00  1660    0.00    0.00    0.00    0.00    0.00    54      23.42
100.00  1991    0.00    0.00    0.00    0.00    0.00    55      28.85
100.00  2129    0.00    0.00    0.00    0.00    0.00    55      31.47
100.00  1712    0.00    0.00    0.00    0.00    0.00    53      24.26
100.00  1609    0.00    0.00    0.00    0.00    0.00    53      22.63

I noticed you are using the acpi-cpufreq CPU scaling driver. I do not know how thermald works with it, but I assume it does.

[andrea.mamprin]

Is this a dual-boot system. Does it still behave normally in Windows if it is?

Have you recently opened the machine and perhaps bumped the cooler?

No my system is not a dual-boot one. Only ubtuntu. A few months ago I was using windows OS and the fans was very noiseless. Then I install ubuntu 16.04 and the fan start work harder. I supposed the problem was due to an excessive use of processor, a batch R script that download json file via Twitter Rest Api is running all the time. I discover that R was creating a giant history archive that blocked my memory. Once this problem has been overcome, the fan was reducing their work just a little bit. Some weeks ago I upgraded Ubuntu (18.04) and for a few days I've not worked with the batch script. Neverthless the fan was cooling almost all the time. So i keep exploring the overheating problem.

I'm not an expert in the hardware part, so this work is done by a computer technician. It was cleaning the pc a few months ago, when ubuntu was already installed. It seems to me that the fan action was the same before and after cleaning.

[CatKiller]

The thermal paste goes between the processor and the heatsink. I'm sure the fans are keeping the heatsink nice and cool, but not actually cooling the processor itself.

Air is a rubbish thermal conductor, so you use paste to keep contact between the chip and heatsink without any bubbles.

[andrea.mamprin]

You can observe the status of thermald via:

doug@s15:~/idle$ sudo systemctl status thermald.service
[sudo] password for doug:
● thermald.service - Thermal Daemon Service
   Loaded: loaded (/lib/systemd/system/thermald.service; enabled; vendor preset: enabled)
   Active: active (running) since Wed 2018-12-19 23:50:55 PST; 2min 52s ago
 Main PID: 13755 (thermald)
    Tasks: 2
   Memory: 1.5M
      CPU: 23ms
   CGroup: /system.slice/thermald.service
           └─13755 /usr/sbin/thermald --no-daemon --dbus-enable

Dec 19 23:50:55 s15 systemd[1]: Starting Thermal Daemon Service...
Dec 19 23:50:55 s15 systemd[1]: Started Thermal Daemon Service.
Dec 19 23:50:55 s15 thermald[13755]: 13 CPUID levels; family:model:stepping 0x6:2a:7 (6:42:7)
Dec 19 23:50:55 s15 thermald[13755]: Polling mode is enabled: 4
Dec 19 23:50:55 s15 thermald[13755]: XML zone: invalid sensor type []

You can re-start it via:

sudo systemctl restart thermald.service

I re-installed it on my test server and set the trigger temperature to 55 degrees. I fully loaded all of my CPUs with thermald disabled, then enabled it. Turbostat output:

doug@s15:~$ sudo turbostat --Summary --quiet --show Busy%,Bzy_MHz,PkgTmp,PkgWatt --interval 15
Busy%   Bzy_MHz PkgTmp  PkgWatt
100.00  3400    66      57.75
100.00  3400    67      58.23
100.00  3400    69      58.59
100.00  3400    70      58.85
100.00  3400    71      58.97
100.00  3199    65      54.54  <<< Thermald service started.
98.96   1678    58      23.68
71.25   1600    55      17.26
69.45   1600    54      16.82
88.33   1600    54      20.33
100.00  1600    52      22.44
100.00  1909    54      27.60
100.00  2114    55      31.15
100.00  1743    56      24.72
100.00  1862    51      26.85
100.00  2045    56      29.91
100.00  2035    53      29.79
100.00  1869    55      26.99
100.00  1945    53      28.20
100.00  1909    54      27.62
100.00  1770    55      25.23
100.00  1817    54      26.08

EDIT: I am using the intel_pstate CPU frequency scaling driver and the powersave governor. For less extreme temperature control thermald limits CPU via the max percent setting:

doug@s15:~/idle$ cat /sys/devices/system/cpu/intel_pstate/max_perf_pct
60

If the temperature is still too high, even at minimum CPU frequency, then thermald starts to use "kidle_inject", which is why you see my "Busy%" drop below 100%, for a short time, above. Also:

doug@s15:~$ sudo turbostat --quiet --Summary --hide Avg_MHz,SMI,GFXMHz,TSC_MHz,GFXWatt,CorWatt,IRQ,POLL,C1,C1E,C3,C6,CPU%c1,CPU%c3,CPU%c6,CPU%c7,CoreTmp,GFX%rc6,Pkg%pc2,Pkg%pc3,Pkg%pc6 --interval 5
Busy%   Bzy_MHz POLL%   C1%     C1E%    C3%     C6%     PkgTmp  PkgWatt
100.00  3400    0.00    0.00    0.00    0.00    0.00    74      59.69
100.00  3400    0.00    0.00    0.00    0.00    0.00    74      59.74
100.00  3400    0.00    0.00    0.00    0.00    0.00    75      59.69
100.00  3400    0.00    0.00    0.00    0.00    0.00    75      59.69
100.00  3400    0.00    0.00    0.00    0.00    0.00    74      59.68
100.00  3400    0.00    0.00    0.00    0.00    0.00    76      59.74
100.00  3400    0.00    0.00    0.00    0.00    0.00    75      59.72
100.00  3400    0.00    0.00    0.00    0.00    0.00    75      59.74
100.00  3241    0.00    0.00    0.00    0.00    0.00    71      56.11   <<< Thermald enabled
100.00  2358    0.00    0.00    0.00    0.00    0.00    65      36.86   <<< Thermald starts by lowering the maximum CPU frequency
100.00  1712    0.00    0.00    0.00    0.00    0.00    62      24.71
99.88   1600    0.00    0.00    0.00    0.00    0.12    61      22.77    <<<< Processor is still too hot, even at minimum CPU frequency. Thermald starts using kidle_inject
92.67   1600    0.00    0.00    0.00    0.00    7.34    60      21.36    <<<< Kidle_inject forces C6, the deepest and lowest energy idle state (for mine, and your, processor)
80.66   1600    0.00    0.00    0.00    0.00    19.35   58      19.07
61.64   1600    0.00    0.00    0.00    0.00    38.38   56      15.49
50.57   1600    0.00    0.00    0.00    0.00    49.45   56      13.51
49.85   1600    0.00    0.00    0.00    0.00    50.17   55      13.35  <<<< Minimum CPU frequency and 50% idle.
49.85   1600    0.00    0.00    0.00    0.00    50.17   55      13.33
49.85   1600    0.00    0.00    0.00    0.00    50.17   53      13.28
49.85   1600    0.00    0.00    0.00    0.00    50.18   52      13.28
49.85   1600    0.00    0.00    0.00    0.00    50.18   52      13.23
51.60   1600    0.00    0.00    0.00    0.00    48.42   51      13.56  <<<< Temperature undershoots target
57.74   1600    0.00    0.00    0.00    0.00    42.28   51      14.71
64.30   1600    0.00    0.00    0.00    0.00    35.71   51      15.86
71.26   1600    0.00    0.00    0.00    0.00    28.75   51      17.17
77.15   1600    0.00    0.00    0.00    0.00    22.86   52      18.18
83.29   1600    0.00    0.00    0.00    0.00    16.72   51      19.30
89.44   1600    0.00    0.00    0.00    0.00    10.57   52      20.44
96.26   1600    0.00    0.00    0.00    0.00    3.74    52      21.70
100.00  1600    0.00    0.00    0.00    0.00    0.00    52      22.38  <<<< Thermald has now removed kidle_inject
100.00  1600    0.00    0.00    0.00    0.00    0.00    52      22.41
100.00  1600    0.00    0.00    0.00    0.00    0.00    52      22.38
100.00  1668    0.00    0.00    0.00    0.00    0.00    53      23.42  <<<< And even starts to increase the CPU max Frequency
100.00  2071    0.00    0.00    0.00    0.00    0.00    56      30.24
100.00  2050    0.00    0.00    0.00    0.00    0.00    53      29.98
100.00  2220    0.00    0.00    0.00    0.00    0.00    55      33.10
100.00  1832    0.00    0.00    0.00    0.00    0.00    53      26.17
100.00  1600    0.00    0.00    0.00    0.00    0.00    52      22.40
100.00  1788    0.00    0.00    0.00    0.00    0.00    55      25.44  <<<< It is not a great servo system, so oscillates some.
100.00  1660    0.00    0.00    0.00    0.00    0.00    54      23.42
100.00  1991    0.00    0.00    0.00    0.00    0.00    55      28.85
100.00  2129    0.00    0.00    0.00    0.00    0.00    55      31.47
100.00  1712    0.00    0.00    0.00    0.00    0.00    53      24.26
100.00  1609    0.00    0.00    0.00    0.00    0.00    53      22.63

I noticed you are using the acpi-cpufreq CPU scaling driver. I do not know how thermald works with it, but I assume it does.

Following your suggestion I've checked if termald was working. The status is active, nevertheless the ouput tells me that "sysfs write failed".

andrea@andreaUbu:~$ sudo systemctl status thermald.service[sudo] password for andrea:
● thermald.service - Thermal Daemon Service
Loaded: loaded (/lib/systemd/system/thermald.service; enabled; vendor preset:
Active: active (running) since Fri 2018-12-21 09:49:43 -05; 1h 36min ago
Main PID: 814 (thermald)
Tasks: 2 (limit: 4915)
CGroup: /system.slice/thermald.service
└─814 /usr/sbin/thermald --no-daemon --dbus-enable


dic 21 09:49:44 andreaUbu thermald[814]: 13 CPUID levels; family:model:stepping
dic 21 09:49:44 andreaUbu thermald[814]: Polling mode is enabled: 4
dic 21 09:49:42 andreaUbu systemd[1]: Starting Thermal Daemon Service...
dic 21 09:49:43 andreaUbu systemd[1]: Started Thermal Daemon Service.
dic 21 09:49:46 andreaUbu thermald[814]: sysfs write failed /sys/devices/virtual
dic 21 11:06:36 andreaUbu thermald[814]: sysfs write failed /sys/devices/virtual
dic 21 11:06:36 andreaUbu thermald[814]: sysfs write failed /sys/devices/virtual
lines 1-15/15 (END)...skipping...
● thermald.service - Thermal Daemon Service
Loaded: loaded (/lib/systemd/system/thermald.service; enabled; vendor preset: enabled)
Active: active (running) since Fri 2018-12-21 09:49:43 -05; 1h 36min ago
Main PID: 814 (thermald)
Tasks: 2 (limit: 4915)
CGroup: /system.slice/thermald.service
└─814 /usr/sbin/thermald --no-daemon --dbus-enable


dic 21 09:49:44 andreaUbu thermald[814]: 13 CPUID levels; family:model:stepping 0x6:3a:9 (6:58:9)
dic 21 09:49:44 andreaUbu thermald[814]: Polling mode is enabled: 4
dic 21 09:49:42 andreaUbu systemd[1]: Starting Thermal Daemon Service...
dic 21 09:49:43 andreaUbu systemd[1]: Started Thermal Daemon Service.
dic 21 09:49:46 andreaUbu thermald[814]: sysfs write failed /sys/devices/virtual/powercap/intel-rapl/intel-rapl:0/enabled
dic 21 11:06:36 andreaUbu thermald[814]: sysfs write failed /sys/devices/virtual/powercap/intel-rapl/intel-rapl:0/enabled
dic 21 11:06:36 andreaUbu thermald[814]: sysfs write failed /sys/devices/virtual/powercap/intel-rapl/intel-rapl:0/constraint_0_power_limit_uw

In this link I found this tip: "This probably means that the system vendor has disabled modifying the power limits for the CPU". Is it right? How can I bypass this situation?

[CatKiller]

Have you got

intel_pstate=enable

in your grub? It used to be a necessary step, but I don't know if it still is.

[andrea.mamprin]

Have you got

intel_pstate=enable

in your grub? It used to be a necessary step, but I don't know if it still is.

Yes I've enabled it and then disabled it. It seems to me that the fan works harder when pstate is enabled.

[Doug S]

In this link I found this tip: "This probably means that the system vendor has disabled modifying the power limits for the CPU". Is it right? How can I bypass this situation?

For my configuration file, you don't need to bypass the situation, because it shouldn't exist. That stuff is disabled for my processor also. (use the intel_pstate CPU frequency scaling driver)

[Doug S]

Have you got

intel_pstate=enable

in your grub? It used to be a necessary step, but I don't know if it still is.

It is no longer needed. It is the default CPU frequency driver for the appropriate Intel processors.

[pqwoerituytrueiwoq]

Unless your ambient temp is near unbearable you should not be able to get temps of 80C+ without a synthetic stress test (or mining software) running without a hardware level issue, bad cooler mount, bad paste application, no airflow
you should make sure the cooler is mounted properly and you have a good paste application
A lot of people flip out about how paste is applied, but given you use a NON-CONDUCTIVE paste, too little paste means overheating, too much makes cleaning it off in the future a mess
*Some CPU coolers come with a peal off plastic cover on the bottom you need to remove before installing
can you take some pictures around the cpu cooler in your system, maybe it is bad enough for us to see something from them


If you see lower temps with another OS this is probably why
The GPU load heating up the cpu, gnome is probably keeping the gpu under enough load it can not idle
if this is correct you will see lower temps using xubuntu or lubuntu, you can test this live

Assuming this is just cause the load is very high cause of the gnome desktop eating the GPU's processing power you could get a older gaming class GPU like the GTX 600 series or RX 400 series or getting a better cpu cooler like a Hyper 212 which run 15 to 30 USD make sure it will fit in your case before buying