X200 lm-sensors *PIC*

[salamisami]

Anyone know which sensor might be for which device?



I deduced that the sensors 5 and 7 must refer to the battery (before labeling them so of course) , since removing the battery makes them N/A

I'm particularly interested about the two sensors reporting the highest temps - temp2 and temp9. They're high compared to the rest and refuse to decrease when the laptop is idling.

As you can see, my system performs very cool otherwise - thanks to linux_phc - but those two keep the builtin hw-fan controller running at over 3k rpm which is quite audible in a quiet setting.

[salamisami]

Laptops been idling on the desk for a hour now, charging - temp2 just gets higher, its at 46 C now! Temp9 is also higher, at 42C.

[tarvoke]

on mine (x200t but ought to be similar), #2 is HDAPS i.e. the HD temp, #4 is GPU, #9 and #10 are basically both pci bus (I would chalk it up to radio temperatures... have you tested what happens when you disable wireless?)

active disk and radios are going to keep the machine warm...

[salamisami]

on mine (x200t but ought to be similar), #2 is HDAPS i.e. the HD temp, #4 is GPU, #9 and #10 are basically both pci bus (I would chalk it up to radio temperatures... have you tested what happens when you disable wireless?)

active disk and radios are going to keep the machine warm...

HDDtemp reports my SSD to average around 33 degrees C. Disabling wireless made no difference, apparently there's no such sensor.

[tarvoke]

HDDtemp reports my SSD to average around 33 degrees C. Disabling wireless made no difference, apparently there's no such sensor.

there's a lot to consider and more than a little nonsense going on inside the machine -

- many of the sensors do not give truly meaningful/honest output, sometimes this is simply needing calibration/interpretation, sometimes purely wrong and can't be helped. the CPU core temps are the most trustworthy (modulo if the coretemp driver has a decent idea of what the junction limit should be - often, it doesn't). the HDD temp "should" be reliable.

- there is still some voodoo on the architecture, lines like PROCHOT# and THERMTRIP# which can assert and cause P-state or C-state changes or HLT/poweroff without even letting software/OS knowing about it. (and we all need to thank our lucky stars for this voodoo, this is what saves your CPU from frying to death regardless of software)

- there is ACPI stuff going on, unrelated to the above-mentioned 2 registers, unrelated (except not exactly) to any of the stuff showing up in sysfs hwmon/libsensors arena. but you can see the ACPI temps are more or less identical to the ISA temps.

- it's not exactly a closed system, but still, the temperatures of one area are going to affect other areas, it starts to approach some overall homogeneous state. charging battery is warm, and will warm other sensors up too. discharging battery is warm, ditto. wifi is warm, will warm up the disk and CPU. everything warms up everything, and the only sinks are either direct radiation or the cpu/hsf.

- this is more anecdata, but my personal experience on 5 different generations of X series machines, if you have 1 or 2 or 3 of the radios enabled, the palm bezel can get pretty warm. I think most people experience this too? and on my crappy x100e, wow, in summertime I would have to turn wifi off, otherwise both cores would overheat and the machine would halt (or sleep, if I was lucky) - but turn wifi/radios off, and the palmrest becomes cool, the CPU doesn't overheat (again, on x100e, this was/is seriously stupid, with wifi the cores can jump to 92dC if you even look at the thing wrong)

ok, TMI, sorry. just... don't expect the sensors to be the most accurate or honest...

[salamisami]

I did not mention earlier, but earlier I had a p8400 motherboard with faulty mpci ports so i replaced with new p8600 motherboard. The OS on the SSD worked just fine without reinstalling anything, but is it possible that it might be causing some of the heat / sensor issues? I had installed phc kernel modules and some such things, maybe there is some conflict with having another cpu, even though they are almost identical and the chipsets are the same?

[tarvoke]

I did see your other thread! which was very interesting reading thank you! (also? great work on the swap!)

(((enough of the phc stuff, both intel and amd, has made it into main kernel, I didn't realize people were continuing using phc - I used to use it exclusively - voltages and c-states - , but is there still a reason to?)))

between the default config which libsensors already installs (acpi / isa coretemp / isa thinkpad), and the coretemp object/driver, you really should not have to do anything extra.
you could run "sudo sensors-detect" but all it will do is say: "blah-di-blah, I have detected ISA coretemp, do you want me to rewrite the conf file?"
but it's not going to calibrate/re-calibrate.

this... gets into some even weirder voodoo.
- Intel doesn't publish certain specs until core devices are EOL.
- Intel cores may or may not contain registers with values for heat limits. these are actually calibration values, not heat limits (did I mention, it gets confusing?)
- if the coretemp object/driver doesn't get any useful input, it sort of just punts/fudges and "usually" does good enough.

knowing the specific junction limit for your CPU can help calibrate, and you can pass the number to your kernel cmdline, or to the coretemp module options
https://www.kernel.org/doc/Documentation/hwmon/coretemp
looking at your sensors output, it looks like it guessed correctly. it's about the same on my x200t.

critical-temp of 105dC, this is around the max junction temp. this is also semi-coincidentally where ACPI would/will have tried to shut things into RAM-sleep. the 127dC (eh, call it 125) is usually where THERMTRIP# kicks in, no user or OS interaction required - the CPU safely halts every-darn-thing.

[salamisami]

On second thought, the sensors might be related to the wifi card after all: I switched power saving mode on wlan2 and the temps seem to have lowered by a few while charging, and by one or two while on battery.
I find this pleasing, do you think switching to another manufacturer's Wi-Fi card might decrease the temps further? I've currently got the Intel 5100 AGN card. I'm aware I'll have to install the whitelist bios hack, nothing I haven't done before.

[salamisami]

Update: Removed the wifi card and booted, checked sensors, temps 2 and 9 are still reporting higher-than average temps... I guess apparently its not related to the wifi card after all, but the sensors are likely to be close to the location, since lowering the power draw did lower the temperatures.

[salamisami]

Hmm, according to http://vminko.org/gentoo_manuals/thinkpad_x200 the temp2 sensors refers to "HDAPS". Is there any way to make those temps lower?