I remember a bug where passive cooling kicked in too late.

I don't have a machine with passive cooling trip point in my hands currently, 
you can debug this
with a command like that:
watch -n1 cat /proc/acpi/thermal_zone/*/{temperature,trip_points,state} 
/sys/devices/system/cpu/cpu0/cpufreq/{scaling_max_freq,scaling_cur_freq}

As soon as passive trip point is exceeded, cpufreq max must be reduced 
immediately and the state in the thermal zone must switch to passive.
There was a bug that this is done in the next measure cycle.
The temperature is checked and adjusted each polling_frequency seconds or if a 
thermal ACPI event happens.
*This is not the case if passive cooling kicks in*. In this case the a BIOS 
value tsp (1/10s) is used which is very high on the ThinkPads (600 per default).
That means if passive cooling kicks in (which is quite normal on latest 
ThinkPads on high load, they reach high frequncies very quickly) the temperature
is only checked every minute as soon as the passive trip point has been 
reached. Reducing max_freq by one freq step should be enough, but that must
happen as soon as passive cooling kicks in.

The tsp value (time in 1/10s how often temp should be polled when
passive cooling is on) can be overridden by passing as thermal module
parameter.

The mentioned fan issue:
Updating the EC firmware might help a bit (on ThinkPads). There was a firmware 
(EC, not BIOS) update stating "avoiding fan noise", IIRC this change got 
reverted
in some later firmware update. But I doubt that is the real problem. In fact, 
all this must work without fans, cpufreq should just go down to lower steps.

Another thing that might help:
If you have an ATI Radeon graphics card, make sure
Options "DynamicClocks" "On"
(-> better check if this is correct) is added in your xorg.conf
I heard this can reduce temperature significantly on some machines.

-- 
CPU overheats during high usage "throttling <not supported>"
https://bugs.launchpad.net/bugs/22336
You received this bug notification because you are a member of Ubuntu
Bugs, which is a direct subscriber.

-- 
ubuntu-bugs mailing list
ubuntu-bugs@lists.ubuntu.com
https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs

Reply via email to