[ltp] T22 temperature

FeRD linux-thinkpad@linux-thinkpad.org
Wed, 18 Oct 2006 15:10:28 -0400


This is in reply to an ancient message, but I'm just catching up on 
reading T2x-related messages in my ltp inbox. Since nobody else seems to 
have replied on this thread I figured I'd chime in, and at least get my 
experiences documented in the hopes they may be of some use. (Plus, my 
T22 is not long for this world, having exceeded its reasonably-useful 
lifetime by at least a year already.)

Crispian Thorne wrote:
> I recently upgraded my venerable T21 800MHz to a 1GHz by way of buying a
> stripped T22 bottom tray on eBay (I had to replace the system board and
> so the upgrade seemed a good idea as well).
>
> Everything seems to work great, but I've noticed that it seems to run
> hotter than the old processor did- my gkrellm temp monitor runs at
> around 120-130 F most of the time, rising to 150-160 F under heavy load.
> This in an air conditioned room at an ambient temp of around 72-75 F.
>   
My T22 (with factory 1GHz PIII) in a "room-temperature" room tends to 
clock in at 55C-60C (100-110F) when in-use but not really "doing much" 
(in other words, like right now, when I'm just typing into Thunderbird 
and mousing around a bit)... as soon as I start doing anything 
CPU-intensive (say, a yum/yumex update or playing video in xine), the 
temp starts flying up into the red. The temps you're seeing under heavy 
load don't sound that unusual to me -- but I've also found they're not 
particularly stable, and that the machine's built-in cooling responses 
aren't enough to keep it from working itself into an overheat-related 
(temporary) failure or instant shutdown.
> Are these temps normal, and does anyone know what the max temp for this
> processor is (Mobile Pentium III)?
>   
I haven't seen Intel or IBM official specs, but clues appear in 
/proc/acpi/thermal_zone/THM0/trip_points:

critical (S5):           97 C
passive:                 92 C: tc1=5 tc2=3 tsp=600 devices=0xdfffc760

...In my T22, without any settings by me, the ACPI thermal monitoring is 
configured to attempt passive cooling at 92C  (165F) and to hard-bail 
into S5 at 97C  (174F).

The machine will definitely run at temps above 90C. However, it will 
also keep getting hotter if running with load and the throttle wide open 
- whatever's supposed to kick in at 92C doesn't seem to help much at all.

Unfortunately (or fortunately, since my CPU still hasn't leaked anything 
unfortunate and continues to function), that 97C critical-shutdown 
marker isn't just for show. If I forget to take some sort of limiting 
action BEFORE I settle in to watch a DVD or a long MPEG file, I'm 
guaranteed not to see the end... unless the movie's final scene really 
is a black screen accompanied by the sound of a HDD spinning down.

Usually, I just try and remember to set the CPU freq down to 700MHz 
(fast enough to do a passable job with most video) before I start any 
long-duration load-producing tasks. At 700MHz, I don't have to worry 
about tripping the critical point -- it barely even gets above 80C, 
running at full load. And whenever I'm running the machine in an 
un-climate-controlled room (like my home) where the ambient air won't 
assist in its cooling, I usually set the CPU throttling 
(/proc/acpi/processor/CPU0/throttling) to T1 for a 12% reduction in CPU 
speed -- and corresponding reduction in heat output even at 1GHz. It's 
enough to give me headroom so I can notice GKrellm strobing temperature 
alerts at me and react to them BEFORE the machine goes dark. (Depending 
how hot the box is, that reaction may be frantically Ctrl-Z'ing or 
killing whatever job is producing the CPU demand, at least for a few 
seconds so the temp can coast down.)

The one remaining thermal problem I've been having is coming from an 
unlikely source -- the HDD. I replaced my original IBM drive with a 60GB 
Samsung drive that tends to (you guessed it) heat up under heavy use. It 
also isn't cooled very well by the system -- if the CPU and BAT temps 
get too high for too long, its own temp creeps up and only giving it 
some idle time will S-L-O-W-L-Y lower the temp.

Problem is, the thing runs at 52-54C minimum, 56-57C as soon as it's 
asked to do anything disk-intensive... and as soon as it gets to 59C, I 
have to get it cooled down within 10-15 minutes or SMART will start 
counting Hardware_ECC_Recovered errors by the thousands, which quickly 
lead to DMA transfer errors on the IDE bus that destroy the system's 
response time, and necessitate a shutdown-and-cool-off period before I 
can continue using the laptop.

So, definitely don't trust the built-in cooling to cool your T22 enough 
to keep it running... not without some additional 
configuration/scripting to react to higher-load/higher-temp situations. 
It's possible fiddling with the trip_points in ACPI would be enough to 
get the laptop handling things itself, but I never understood how to 
write into that file in a meaningful way.

> - --
> Crispian Thorne
> linux@crispian.org
> *linux & lovin' it*
> Linux User # 329400
>