overheating t43p - 91C

[binary.koala]

hi, my t43p 2668-peg was overheating like crazy - i run second life under linux and then:

Code: Select all

danja@t43 ~ $ tail /proc/acpi/ibm/thermal 
temperatures:   89 50 36 91 34 -128 27 -128 48 54 51 -128 -128 -128 -128 -128

so cpu - 89C, gpu - 91C, isn't it insane? of course it's full load to all components, cpu load 100%, intense graphics. fan is full-on, i switched 'tp-fancontrol' off. the machine is brand new (manuf 01-2007, purchased 04-2007).

what i did for now is replaced cpu with Dothan 1.8, 400fsb, now cpu max 65C, gpu 71, which is (in comparison) sounds more acceptable.
any advices? (i'd like to have full power)

Can 5W in power consumption (dothan1 1.8 vs. dothan2 2.1) make such big difference? what did Lenovo designers think when used so hot cpu with such inefficient airflow? i'm really disappointed in Lenovo. But i also think t43p was kind of a white duck - the time of it's development came to the exact moment of the transition of the line from IBM to Lenovo.

[richk]

The machines had been made by Lenovo in China for a long time before the T43 came out

[_Jake_]

How do you know that you don't have a fan fail? 91C seems too high to be the normal range.

[binary.koala]

How do you know that you don't have a fan fail?
91C seems too high to be the normal range.

well, first of all i can hear the fan, it's notoriously annoying, then i can read rpm and even span though all the speeds via /sys interface in linux. so fan is working, i'm sure (i saw it spinning too!). It blows out hell hot air btw.

[Troels]

A few suggestions, but as i don't have that much experience with any linux distros i'm not too sure;

- Are you sure that the SpeedStepping features of the CPU is enabled? As a default, i think it will run at full clock speed which wouldn't exactly help matters.

The Pentium M Dothan "Sonoma" (2nd gen. Dothan) was and is more inefficient than the Dothan 400 Mhz fsb when it comes to heat. (TDP of 27W vs. 21W)

ATi's powerplay features are probably not enabled in their linux drivers. I'm not sure. Xorg?

Other than that i'm running a T42 with a MR9600. When playing intensive GPU games, it does reach 95-97 deg. C. I don't bother too much with it actually. Surely it's life will be shorter, but when it dies, surely something else wouldn't work. Thereby not saying that the temps are acceptable.

[binary.koala]

- Are you sure that the SpeedStepping features of the CPU is enabled? As a default, i think it will run at full clock speed which wouldn't exactly help matters.

yeah it is on, i also have 'ondemand' governor that lowers the freq of my CPU as needed.

The Pentium M Dothan "Sonoma" (2nd gen. Dothan) was and is more inefficient than the Dothan 400 Mhz fsb when it comes to heat. (TDP of 27W vs. 21W)

27w vs. 21w means that dothan2 is hotter, true. TDP is how much power output (read heat) is being generated by the cpu at top load. unless Dothan2 cpu has _proportionally_ bigger cooler - it _will_ get hotter.

ATi's powerplay features are probably not enabled in their linux drivers. I'm not sure. Xorg?

i'm using xorg radion driver and it has DynamicClock option which i have on, idle GPU is 45-50C which is normal.

Other than that i'm running a T42 with a MR9600. When playing intensive GPU games, it does reach 95-97 deg. C. I don't bother too much with it actually. Surely it's life will be shorter, but when it dies, surely something else wouldn't work. Thereby not saying that the temps are acceptable.

after i swapped the CPU and installed additional aluminum plate between the GPU and the heatsink my top temperatures became more humane - cpu max 69C, gpu max 76C.

just to make things clear i must say that temperature situation is just the same in linux and windows (i of course kept original XP on my t43, enjoying free games microsoft gives to their legal customers that i suddenly became, oOo ).

perhaps i shall just sell Sonoma CPU that i got left and be happy with extra bucks and colder computer.

p.s. i'm gonna undervolt CPU one day, that shall help too.

[FTC]

Hi, don't trash the sonoma core !. There is no real reason for it to run hotter. .. at the same speed, so you can verify by downclocking it (i.e by reducing the max. mutiplier with nhc) that it should produce about the same heat at 1.8Ghz than your 1.8 dothan core.
My guess is that if you mount correctly the heatsing with good thermal paste, (the aluminium plate is for sure a good idea), you can control the sonoma temps... also undervolting helps a lot.

Now, a further idea I just developed in my T42 is adding an aluminium plate ON TOP of the heatsink with some thermal paste (I also add a strip of thermal tape in the sides so that I am sure it does not move) so that I can force contact between the keyboard metal plate and the heatsink. works wonders.. it decreases further 5-6 degrees the CPU temps.

With all of this techniques I've been able to reduce my T42 1.8 100% CPU temps to just under 60oC (note that I also have a HP NX6110 running a sonoma core at almost same temps).

[agarza]

FTC, I need your advice:
I'd like to lower my T42p temps at load (if the CPU is at 100% with stock voltage (1.340V) it would reach about 72-76C!, I applied Zalman STG-1 on the CPU core and the heatsink (as stated on Zalman's instructions. The GPU would reach about 94C when gaming with default clocking values.

What are your suggestions to lower the CPU/GPU temps (GPU also have Zalman STG-1 on it). Could you elaborate more on the mod you did for your machine (with pics maybe). Thanks

[FTC]

Hi,

First of all, your CPU load temps are fine. If you attempt to lower them more you can, but it is 'for the fun of it'. 72-76 C load for one of these CPUs (I assume high end, 2.1Ghz or so dothan / sonoma) is the design temp. Now, your GPU is a bit on the hot side for my liking, but it's not THAT hot (note that load temps for firegls and 9600 are close to 85 - 90 C normally).

Now, these are the techniques you can follow to lower temps :

1. Undervolt your CPU. recommendation : use NHC. Search the forum to learn how to do it. It is quite easy to do, but leave a little margin after you found your operating voltage. This alone will gain you some degrees.

2. Reseat heatsink using good thermal paste for the CPU. For the GPU, my recommendation is to remove the pink and black 'gunk' altogether, use a layer of good quality thermal paste and put two thin strips of thermal tape in the sides of the GPU 'square so that you grant contact of the heatsink to the GPU. (the idea is that you should be able to later remove it if needed). You can also 'fill the gap' with an aluminium or copper sheet and use only thermal paste then (it will be held in place by the pressure of the heatsink).

3. Use your own 'fan control'. My recommendation is to use NHC w/the proper ACPI module and make the FAN run at hi (7) when temp is more than 75C. (and maybe value '64' if it goes over 87 C or so)

4. My previously commented technique of 'forcing contact of the heatsink with the backplate of the keyboard by adding an aluminium piece on top with thermal paste also works wonders. (note that you will 'feel' all your keyboard radiates a bit of heat afterwards, but the CPU will be much cooler). (**)

5. Very important, keep your sink clean of dust.. every 2-3 months clean it up.

.. the following you can do also but will impact your system's performance :

6. Underclock your CPU, or force it to a given multiplier, lower than the max. supported...

7. use powerplay even if on AC power. This will effectively underclock the GPU.

===========================
(**)
.. sorry, I don't have pictures of the process to do my fan mod and am not willing to 'remake' the thing just to take some, but note the diagram :

KEYBOARD <Keys
============== <aluminium plate
<Thermal paste
NNNNNNN < Aluminium thin block
<Thermal paste + two strips of thermal tape
FFFFFFFFFF < Fan


Luckily I found one image of the thing right before adding the last
layer of thermal paste and remounting the keyboard :

[agarza]

Thanks for your suggestions.

Later the next week I will try to test with Arctic Silver 5. however I've found a Web site that tested 36 thermal pastes: it seems the winner is a paste called Arctic Cooling MX-1.
Here's the link:

http://www.xtremesystems.org/forums/sho ... p?t=135359

I should give it a shot. To my liking I think the fan is not blowing all the heat efficiently, this using Zalman STG-1 thermal grease.

[binary.koala]

just to keep you posted guys,
today i put back my original cpu (the one that was overheating, sonoma 2.1, 533fsb) with the intention to give it another shot and refine my cooling system. what i had in mind is that my main problem is GPU and since the heat sink is shared it often heats up the CPU, even when the latter is not under any load. so what i did was bigger/better aluminum sheet on GPU and better paste. unfortunately i didn't have and thermal tape (where do you get it in europe?), so keyboard is not thermally connected yet...
i undervolted my cpu, the rates are here, same applies to linux phc of any windows solution - http://gentoo-wiki.com/HOWTO_Undervolt_ ... e_voltages

now my CPU T doesn't go above 73C* at the max load! great! i got the power back with the same heat output.
the only disappointment is that on lowest freq. the temp is still quite hight - 55-56 C* which is kind of bothering me (and birning my laps).

perhaps i should search or make another thread for this question, but i was just wordering about you average runtimes on battery?
mine is holding 2-2.5 hours at most, how are yours doing?

[agarza]

binary.koala: Good that you lowered your temps.

I also reapplied thermal grease after 6 months without cleaning the fan. I used Arctic Silver 5 on both CPU/GPU. The way I did to put the goop in the CPU was to put half-rice-sized amount on the die of the CPU without spreading it. In the GPU I did spread it but I didn't used any thermal tape (don't know where to find it here in México). I make sure the heatsink is touching the GPU core so no problem there. Then when the heatsink was being installed I twisted it so that any air bubbles trapped between the 2 interfaces weren't there.

I then remove the heatsink to check if the amount of thermal goop in the CPU was enough, it was OK, reinstalled heatsink twisting again.

The laptop now at full load the CPU gets about 71C (1.340V), GPU 91C. If I undervolt the CPU to (1.052V) the max temp is 60-61C.

Now to the battery life issue you mentioned: I think you should underclock the GPU to attain better results in your battery. Typically I would get about 3.5-4 hours (depending on the things I'm doing on my machine).

I found out the values posted in the webpage you post are fairtly inaccurate: If I use 0.700V in the 6x Multiplier, my laptop would freeze, so I just had to go to the next value in the NHC chart to make sure the machine is stable.

Referring to FTC advice on the aluminum sheet on top of the heatsink: I WOULD NOT RECOMMEND IT. I did a variation of that using the thermal pad that came by default on top of the heatsink area just above where the CPU is located. I put some Arctic Silver 5 on the thermal pad and when I stressed out the CPU & GPU the x, c, ss where like stickied, this was because the heat transferred by the thermal pad make the little rubber in the keyboard keys to kind of temporarily melt causing that when you press them, the character keeps appearing in the screen because literally the key is still being pressed. I simply had to clean the heatsink thermal paste.

[FTC]

I put some Arctic Silver 5 on the thermal pad and when I stressed out the CPU & GPU the x, c, ss where like stickied, this was because the heat transferred by the thermal pad make the little rubber in the keyboard keys to kind of temporarily melt causing that when you press them, the character keeps appearing in the screen because literally the key is still being pressed. I simply had to clean the heatsink thermal paste.

Xtr,

I don't belive the 'melting' you comment. Most probably what happened is that somehow you 'shorted' the keys circuit with the Arctic silver, or something similar... In my case you can see that the surface area of the aluminium pieces is quite big, and also something I did not mention before... I used white thermal paste for this application... (the cheap one). I did not want to use arctic silver because of the price, being the surface so big... it could end up being a very expensive mod. Now it looks this could have saved me because of the lack of conductivity. Anyhow, ... note that if my max *load* temps on the CPU are 65C now, then the keyboard plate is for sure not higher than ... 55-60 C ? so it is *very unlikely* that this temp melts anything at all.

[agarza]

Well, the rubber where the keys mechanism is located didn't literally melted, but when I pressed the x, c or s keys the key would keep pressed down a few seconds, like it were glued. The next day the keys were working fine. however I didn't notice a significant temp drop, however you're using aluminum sheets, I didn't.

[FTC]

Yep, maybe this is the reason... I guess arctic silver on top of the thermal gum-like pad becomes quite sticky for the keys... where a bit of white thermal compound over a rigid aluminium surface can't be noticed at all.

Regards