[ltp] ThinkPad T520 overheating

[Martin Steigerwald]

Hello!

What you are best advices for a 4 year old T520 overheating like this:

Apr 24 21:31:30 merkaba kernel: [145233.995957] CPU0: Package temperatu=
re=20
above threshold, cpu clock throttled (total events =3D 3937756)
Apr 24 21:31:30 merkaba kernel: [145233.995959] CPU2: Package temperatu=
re=20
above threshold, cpu clock throttled (total events =3D 3937733)
Apr 24 21:31:30 merkaba kernel: [145233.995961] CPU1: Package temperatu=
re=20
above threshold, cpu clock throttled (total events =3D 3937760)
Apr 24 21:31:30 merkaba kernel: [145233.995964] CPU3: Package temperatu=
re=20
above threshold, cpu clock throttled (total events =3D 3937753)
Apr 24 21:31:30 merkaba kernel: [145234.000962] CPU2: Package=20
temperature/speed normal
Apr 24 21:31:30 merkaba kernel: [145234.000965] CPU0: Package=20
temperature/speed normal
Apr 24 21:31:30 merkaba kernel: [145234.000966] CPU1: Package=20
temperature/speed normal
Apr 24 21:31:30 merkaba kernel: [145234.000968] CPU3: Package=20
temperature/speed normal
Apr 24 21:31:30 merkaba kernel: [145234.431470] CPU2: Core temperature=20=

above threshold, cpu clock throttled (total events =3D 1247442)
Apr 24 21:31:30 merkaba kernel: [145234.431475] CPU3: Core temperature=20=

above threshold, cpu clock throttled (total events =3D 1247461)
Apr 24 21:31:30 merkaba kernel: [145234.432440] CPU2: Core=20
temperature/speed normal
Apr 24 21:31:30 merkaba kernel: [145234.432441] CPU3: Core=20
temperature/speed normal

on while playing PlaneShift or openmw, but also in other occurences of=20=

higher load, such as Akonadi bursting the machine or kernel compile whi=
le=20
apt-get upgrading and things like that.

With sensors I see it at 98 degrees then and 3500 to 3600 rpm fan speed=
.

The machine basically crawls to a halt.


On searching on the net I found various things:

1) It seems to be a know problem with ThinkPads from that time.

2)  Some say its a software issue with Linux thinkpad acpi not using=20=

maximum fan speeds, while Windows uses higher fan speeds. Some people=20=

advice to set it to use higher fan levels, but when buying 3500-3600 rp=
m=20
has been more than enough to allow it to turboboost two cores to 3,2 GH=
z=20
for more than half an hour.

3) Someone suggested the thermal compound component used in these lapto=
ps=20
was crap, and suggested cleaning fan and using a good thermal compound=20=

component.

4) Someone suggest cleaning the fan with a can of air.

Just to name a few suggestions.


What steps would you recommend?

Laptop may be replaced by a newer ThinkPad model as main laptop. The=20=

laptop doesn=C2=B4t shut down=E2=80=A6 so the throttling works, but=E2=80=
=A6 it crawls the=20
machine to a halt.  So I would like to avoid any of the risky stuff tha=
t=20
may brick the machine. Slow is still better than bricked.

BIOS is 8AET63WW (1.43 ) which was quite recent last time I looked.


My observation is that laptop generally runs hotter and fan is on more=20=

often, which is why I think that cleaning or even replacing the fan or=20=

renewing the thermal compound might be best. Interim solution may be to=
=20
set fan speeds higher, but I see this more as a work-around that may we=
ar=20
the fan out even more=E2=80=A6 cause I have never seen the fan higher t=
han 3600 or=20
maybe 3800rpm.

I already set the BIOS "Adaptive Powermanagement on AC" from "Performan=
ce"=20
to "Balanced" in the hope that it may help a bit. Read so. I also appli=
ed:

pcie_aspm=3Dforce i915.i915_enable_rc6=3D7

to kernel command line for testing, but I think that 1) Intel gfx drive=
rs=20
goes into rc6 sleep state by default meanwhile and 2) pcie_aspm issues=20=

have been fixed in Linux long time ago (currently running 4.0 kernel).

This all started quite some time ago, but was only in summer, now its=20=

sufficient to just have the room one or two degree warmer than usually =
(21=20
to 22 degree celsius instead of 20) in order to trigger the behavior, o=
r=20
it just got so worse that its always triggered.

As written, this laptop may be replaced, but I think its a to let a lap=
top=20
which costed 1800 euro degenerate like that, so I want to give it some=20=

attempt to fix it.


I also tried Intels thermald with loads powerclamp and uses intel RAPL,=
=20
but that totally crawls the machine to a halt. Its much worse than with=
out=20
it. It injects idle events up to a point where the machine doesn=C2=B4t=
 do much=20
at all anymore.

Ciao,
--=20
Martin 'Helios' Steigerwald - http://www.Lichtvoll.de
GPG: 03B0 0D6C 0040 0710 4AFA  B82F 991B EAAC A599 84C7