Ubuntu – Laptop CPU+GPU overheating after update to 18.04 LTS

18.04cpu loadinteltemperature

I had this same issue in my last attempt to upgrade (with a clean install) from 16.04 LTS to 18.04 LTS. When I am browsing internet and specially when I am playing videos my CPU+GPU raises to fast to high levels causing throttling to fire and producing intolerable lags, leaving the system unusable for about 60 seconds or more.

With Ubuntu 16.04 LTS I never experienced this problem. Of course temperatures raised and throttling fired frequently (after all this is a laptop), but system was never unusable and video playback was smooth.

My previous solution was to downgrade to 16.04 LTS again, but now I need the newer version so I want to find a solution.

System specs:

CPU: Intel Core i7-6500U
GPU: Intel HD Graphics 520
Ubuntu: 18.04 LTS
Linux: 4.15.0-30
Mesa: 18.0.5
Hardware acceleration: on

UPDATE 1

Definitively it is a problem with frequency throttling. Instead of throttling CPU frequency to avoid overheating the procesor, the system is waiting until temperature reaches a limit (default of 85ºC) and then intel_powerclamp module fires, turning the system unusable for 60 seconds or more.

I would expect a progressive performance degradation from intel_powerclamp, but no, it is all or nothing. ¿Can this be configured?

My current solution is a modification of the script of this answer. It seems to work, but it is a big hack. There must be a conventional way to tweak system configuration to accomplish the same result.

UPDATE 2

If I turn off frequency turbo boost (echo "1" | sudo tee /sys/devices/system/cpu/intel_pstate/no_turbo) the problem completely disappears. Temperature never reaches the limit and the script introduced on update 1 nor intel_powerclamp fires and the laptop is smooth and usable.

Problem now is that turbo boost is disabled. With Ubuntu 16.04 LTS I never had this issue.

This is clearly a bad configuration of the thermal management subsystem of Ubuntu 18.04 LTS. Is there a way to fix it?

UPDATE 3

System is still unusable. Installing Ubuntu 18.04 LTS server and anything else than Gnome3 works. Well, it's not a novelty that Gnome sucks, but it was usable with 16.04…

Best Answer

Disable intel_pstate at boot by editing your grub configuration.

Open /etc/default/grub with your preferred text editor and locate the line that begins with:

GRUB_CMDLINE_LINUX_DEFAULT

And add intel_pstate=disable at the end of that string. Here's how mine looks after that:

GRUB_CMDLINE_LINUX_DEFAULT="quiet splash intel_pstate=disable"

Now, execute:

sudo update-grub

And reboot your laptop. That's it. It should work as smooth and cool as with 16.04.

Some additional information: intel_pstate governor seems to be buggy, or just plain stupid and unable to detect quick temperature raises and reduce the processor speed in consequence. It seems it'd just pray to Saint Cooling Device for it to keep the CPU below the shutdown threshold. In a desktop computer with a DC9 turbine like fan such leap of faith maybe justified, but in laptops with one small fan like my XPS13 it's better to be skeptic and leave those things to the good old ACPI (Atheist Configuration and Power Interface). At least until the intel_pstate governor grow some brains.

And the turbo boost is mostly useless. Chances are that you won't lose anything for having it disabled. That feature was useful in the days of monocore 32 bit processor, but not in the 64 bit multicore world we're living on right now.

Related Solutions

Ubuntu – CPU Turbo boost: modify temperature limit under 14.04

That's not a bug, it's a feature! Check your temperatures, the only reason the CPU speed is decreased is because you're overheating. Look at the first output you show, you have one of your cores running at 92°C! That's pretty hot to begin with and it probably went even higher before your clock speed was reduced.

You can check the maximum temperature your CPU can deal with by running sensors. For example, for my Intel i7:

$ sensors
coretemp-isa-0000
Adapter: ISA adapter
Core 0:       +77.0°C  (high = +95.0°C, crit = +105.0°C)
Core 2:       +79.0°C  (high = +95.0°C, crit = +105.0°C)

So, on my machine, 95.0°C is considered high and 105°C is the critical temperature at which the machine will be shut down. Your specs might be a little different but 92.0°C is definitely high.

Now, you can force your CPU to keep the same clock speed. This is controlled by the CPU governor. You probably have it set to ondemand, but the following options are available:

Performance keeps the CPU at the highest possible frequency
Powersave keeps the CPU at the lowest possible frequency
Userspace exports the available frequency information to the user level (through the /sys file system) and permits user-space control of the CPU frequency
Ondemand scales the CPU frequencies according to the CPU usage (like does the userspace frequency scaling daemons, but in kernel)
Conservative acts like the ondemand but increases frequency step by step

To change your governor to, for example, "Performance", run this:

echo "performance" | sudo tee /sys/devices/system/cpu/cpu0/cpufreq/scaling_governor

However, the likely result will simply be that your computer will shutdown when the high temperatures are reached. It might also still scale it down, I'm not sure how the safety features are implemented. In any case, the main point is that you don't want to do this because it can seriously harm your hardware.

There are very good reasons why your computer won't let you go past certain temperatures and you don't want to screw with that. I don't know of any way to actually disable the temperature limit and I wouldn't tell you if I did. There are easier ways, but disabling the temperature safety limits is certainly a good way to destroy your computer.

Ubuntu – Intel Turbo boost on ubuntu 14.04

I don't think you know what you are talking about, because you describe something that is very different from how this technology was initially supposed to work.

Turbo Boost, glitter for standard clock speed

Turbo boost will not automatically overclock your CPU if both of your cores are already constantly at 80-100% load. Some later variations (depending on the CPU model) will boost under full load for very short intervals to then cool down with negligible long term performance impact. All in all you won't achieve a constant overclocking boost, because as I said, this is not how this feature is designed.

Themal design and cooling

Also one should first provide a better cooling solution before trying to mess with the thermal thresholds that the system designers and engineers carefully defined during the product development. In general, you should never even think about modifying temperature limit for laptops (as the other question initially asked for).

Possible traps: the wrong monitoring tool

There was some confusion when Turbo Boost was introduced, because some tools that monitored the frequency and/or changed governors interfered with Turbo Boost and prevented it from properly working.

Some hardware monitoring applications or panel-applets that allow to display or modify processor clock speeds bring disorder into the fragile arrangement, hence one should stay away from these when in doubt [if turbo boost is working correctly].

Original citation in German:

Einige Hardware-Monitoring-Programme oder Panel-Applets zur Anzeige oder Konfiguration der Prozessor-Taktstufen bringen dies fragile Gefüge aber durcheinander, daher lässt man von diesen im Zweifel besser die Finger.

^{Heise: Turbo Boost unter Linux (probably better known for its English site The H)}

Conlusion

While you can try the link (which is not about Turbo Boost but P-States, which is more technical and about power management) posted in an answer on the other question, you have to realize Turbo Boost is not about constant overclocking under constant heavy load, but improving performance for short bursts, which won't reduce workloads like video editing significantly.

Also you have to realize that marketing (deliberately conf-)uses higher specs that are only achieved in certain boost conditions (or sometimes never, like on some ultrabooks), instead of the standard clock speeds. (This is not very new, the Pentium 4 processors have been known for clocking lower due to overheating, it just got more complicated and made its way to graphics cards, tablet and smartphone processors.)

Simply speaking, if you bought a product with a 2 core CPU at a clock speed of 2.4 GHz and thought you get a 4 core CPU with 3 GHz clock speed or something in between, then realize that you have been fooled or the one who sold you the product didn't know any better.