Re: Recurrent alerts "Package temperature above threshold, cpu clock throttled"

[<l0f4r0@tuta.io>]

Hi Nicholas, all,

Thank you very much for your answer and sorry for my delay...

6 janv. 2020 à 05:33 de nsteeves@gmail.com:

> While I'm very much a junior member of the team, I have noted similar
> behaviour on a laptop as old as an X220 and wonder if it might be due to
> Spectre mitigations.  I have not noticed such log entries after
> disabling hyperthreading.  FYI, as someone who works with realtime audio
> I've found that hyperthreading is detrimental to worst-case latency.
> eg: after many hours trying to workaround troublesome unpredictable
> latency spikes the solution to disable hyperthreading emerged as the
> simplest solution.
>
I've just deactivated hyperthreading in my UEFI.
Still the same warnings in journalctl :(

At least, Internet resources indicate that's it's safer this way with HT deactivated (regarding MDS attacks) but I don't know (and really not anyone as it really depends on current CPU tasks) how my performances are impacted now...
In your case, it seems to be a benefit but surely it's not always the case...
Maybe I can just continue like this and restore HT if this is impossible to live with ;)

>>
>>
> Have you
> tried disabling CPU freq boost?  When the ambient temperature is above
> 27°C my X220 and X230 need to have boost disabled to avoid
> overheating/throttling.
>
I've seen at least 2 ways to deactivate turbo boost:
1) echo "1" to /sys/devices/system/cpu/intel_pstate/no_turbo
Visibly, as sysctl only works with /proc/sys (and not /sys), this needs to be set permanently via a systemd service. What do you think about this procedure: https://blog.christophersmart.com/2017/02/08/manage-intel-turbo-boost-with-systemd/?
2) modify MSR registers via wrmsr (https://askubuntu.com/a/619881). I don't know if there is persistance here...

Do you use any of these? Something else?

PS: My CPU is Intel(R) Core(TM) i7-8565U CPU @ 1.80GHz.
PPS: Even if it works, I don't know if I'm not gonna lose some more performances...

Actually my main questions are: 

i) Are those temperature warnings legitimate?
I mean maybe those warnings have been wrongly triggered because probes are not accurate? I'm not speaking about HW failure (my laptop is brand new) but maybe I just don't have the right driver somewhere...
Indeed, it would be too bad to decrease all my performances with HT/turbo-boost disabled (security apart) whereas the journalctl warnings are wrong initially... ;)

PS: I have installed intel-microcode 3.20191115.2~deb10u1. Still the same issues.
I don't know if it's an issue but microcode module is blacklisted in /etc/modprobe.d/intel-microcode-blacklist.conf (it seems to be a precaution regarding unsafe updates).

ii) Is it risky to do nothing about these temperature warnings?
I have no idea what EC means (Embedded Controller?) but you said EC eventually shutdowns the laptop if need be. I presume it's not really beneficial from the user point of view as the current tasks will be shutdowned and some work/data might be lost during the process.

Thanks again & Best regards :)
l0f4r0