A31 - CPU Overheating and Causing Shutdown

[BuddhaNature]

Hello

I have an A31-2652-Q3C Thinkpad bought new in 2003. It has served very well and I do want to hold on to it as other than this problem it is in very good condition.

Recently I started to experience strange shut-downs. At first I could not get a handle on what was causing the problem and suspected some unknown software issue. Completely re-installed the system but ended up with same result - system would shut-down inexplicably.

Eventually I traced it down to that the CPU when under heavy load is overheating and the system is going into emergency shut-down to protect the CPU.

What do I do now? If it involves replacing the fan what alternatives do I have? Are there fans that other A31 users have installed that are known to be better than that which is shipped with an out of the box A31? If there are better alternative fans, can they be installed without involving modifications to anything (i.e. slot out the old, slot in the new - all done)?

Any advice along these lines would be greatly appreciated.

[Harryc]

If it were my machine I'd replace the CPU and the fan. The stock fan should be fine and I don't recall the P4 M CPU's being terribly expensive...so it makes sense to replace both.

[BuddhaNature]

Hi Harry,

Thanks for the reply. You might not believe this, I barely believe it myself, and now I feel like renaming this post to 'let this be a lesson to everyone - especially me'. Here's what I did:

About half an hour after posting the message I decided to open the Thinkpad and re-new the thermal grease compound at the top of the CPU and give the fan unit a clean. I did just that. However, as I reckoned I had nothing to lose, I took the whole fan assembly, dunked the lot in hot detergent water and give it a good scrub, copper plate, blades and all. Having done that I rinsed it all under a strong jet of hot water from the tap and was quite amazed to see, still, a fair amount of collected dust come out of the unit even though I thought I had just cleaned it thoroughly. Then I put it on top of a radiator to dry while I cleaned all the previous compound-residue off the CPU die. I didn't go so far as to chuck it in the water, but I sure gave it a good scrub with a nail brush. After that I re-did the thermal grease, mostly according to the instructions by 'Raddish' here:

http://www.thedailydeals.com/forums/arc ... 14075.html

I didn't have any lab alcohol so when I cleaned the copper plate I just used one of those 'Scotch Brite' abrasive pads. (I don't think I would do that again as it did scratch the copper a bit but it also really did get rid of all the previous compound residue which, on the plate I had, was like a stamped-on residue that wouldn't budge with just rubbing at it; I doubt alcohol would have budged it either.) To replace the thermal compound I should mention that I didn't have 'Arctic Silver', just a bog standard compound bought out of the local electronics shop about a year ago.

Then I put the machine back together and started the system, hoping that I wasn't about to see my CPU putter a few shots of smoke and die before it even got to Windows.

I have MobileMeter set to auto-start on boot so I thought I might get a quick glimpse of anything seriously untoward, if it got that far. Well it started fine and I was truly amazed to see the start temperature, as the system got into Windows, at around 40oC! The reason why I'm 'exclamation marking' the temperature is that never in the years I've had this computer have I ever seen it run at 40oC. The most usual 'start' temperature I've seen on this system is around 60oC. And that is across two different CPUs that I've had installed in it.

Well the temperature did, of course, go up. But the difference is very dramatic. Under the same general [for me normal] CPU load conditions to which this system is usually subject the average CPU running temperature is now lower by 30oC. No, that's not a typo. Thirty degrees centigrade cut off the operating temperature of this CPU under my normal load on it! You should see the difference.

Since the day I've had this laptop it has 'speed-stepped' its way from either 2.00GHz or 2.50GHz (my first processor was 2.00GHz) to 1.20GHz as regular as clockwork and all in response to the operating temperature of the CPU. Even though I thought that was the system's normal (i.e. correct functioning) I was finding that frustrating to watch. After all, wasn't it a 2.00/2.50GHz CPU I had? Well, yes. But only half the time, the other half it was 1.20GHz. Under sustained heavy CPU usage it would spend significantly more time at 1.20GHz than at 2.00GHz/2.50GHz.

Now that has been significantly altered. I left the system to run overnight but set things up so that the CPU was running at 100% load at all times. When I got up this morning it seems that the CPU has not dropped speed at all at any point to 1.20GHz. It now runs at 2.50GHz permanently and with the CPU temperature topping out and levelling around 66oC-69oC. This system when it is now under sustained full load of the CPU is running around 15oC-20oC cooler than it was before. I didn't even know this system is capable of doing that. It literally feels like I've got a new computer and one that, effectively, runs at just over twice the rate of the one I had before and at significantly lower temperatures.

If I try to account for what has happened here then I think the following is most likely the case.

1. The problem with the enforced thermal shut-down was caused by the thermal compound having recently shrank, crept, back off the top of the CPU die over time – that was fairly evident when I opened the unit up.

2. The original CPU in this Thinkpad was 2.00GHz. This is a CPU that was factory fitted by IBM. It had an operating behaviour as indicated above – speed-stepping all the time between 2.00/1.20GHz. The thermal trigger for the step occurred at around 87oC – 89oC. This trigger value was arrived at with great regularity.

3. About a year ago I fitted a second-hand 2.50GHz CPU. However, I had never fitted a CPU before and looked on the net for instructions. I read about applying the thermal compound. Nevertheless, I wasn't confident about the advice to 'spread the thermal compound thin' and (on that occasion) spread it to accord with what looked like the amount of compound that had been applied to the factory fitted 2.00GHz CPU and in the manner in which if looked like it had been applied. The result of doing this to the 2.50GHz CPU was to set up conditions such that it operated as the 2.00GHz had done – speed-stepping its way down to 1.20GHz all the time in response to thermal protection issues of the CPU. Again, thermal trigger for the step occurred at around 87oC – 89oC. Again, this trigger value was arrived at with great regularity.

4. This time I did in fact spread the compound thin (a layer so thin it was close to transparent) on top of the die and took care to wipe off any compound that had accidentally curled over onto the side of the die. I only had tissues available to do the wiping off both at the sides of the die itself and across the entire copper-plate. The IBM fitted CPU had thermal compound not only on top of the die but around the sides of it as well.

And that is that. Out of this seeming disaster I now have a Thinkpad that, when running on mains power, never speed-steps and runs at temperatures lower than I have ever seen before on this system. And, I have gained that without even following to-the-letter best procedure for fitting a CPU. What all this means, of course, is that for those Thinkpad owners out there that have a machine that is speed-stepping all the time while on mains power, it might very well be worth while opening the machine and having a go at dealing with cleaning the fan and replacing the thermal compound that IBM fitted. It seems that there is no guarantee that IBM's factory will necessarily have got it right.

As a matter of curiosity – even of future assistance – over this kind of thing it would be nice if some folks reading this would comment on how their Thinkpads behave in this regard.

In any case, I hope this post helps someone else sometime. I still can't believe it. 30oC drop in operating temperature under my normal working load on the system. That is something to write about.

Thanks very much for your response, Harry.

I'm happy to say I didn't get that far – this time.


P.S. I'm not an expert on these matters so if I've made any serious error I'd be grateful if those more knowledgeable would correct.

[Harryc]

I could pick out a few things that I would not have done (.e.g. 'fan in water' and scotch brites ), but hey it works and that is the important thing. Enjoy your newly rejuvenated machine.

[BuddhaNature]

I agree on the 'Scotch Brite'. Totally reprehensible on my part. But at two o'clock in the morning the devil was in me for a fix. (Know what I mean?)

However, for the dunking in water. Well... I've found that if you're in a 'prepared to lose, though you might win', situation (I did have nothing to lose) then it is worth a shot.

Over the years, to name a few items, I've had a radio, an alarm clock, a hi-fi unit, a television, a hoover and two entire Atari STe's all hit the tub. On every occasion it has effected the fix (doesn't mean that on all future occasions it will work, of course). Though I think it important to mention that after a swim in detergent they need a good sloshing around in fresh. Also, they need to be thoroughly dried in, say, the hot-water tank cupboard (closet in American, I think). For a television that would mean a couple of days drying-off - just to be on the safe side.

I wouldn't advocate the treatment as a pure matter of course, but I'm not the only one that thinks like this. Try a search in Langalist

http://www.langa.com/

and you can see that Fred shares the same philosophy: If you're in a press and have nothing to lose, well... Splash!

Just dry it out well afterwards.

[Harryc]

There's two problems with using water on a mechanical assembly, like a fan. The first you already mentioned, it must be thoroughly dried out afterward or there is a very real possibility of short circuiting the electronics in it. The second problem is that water will wash away any lubricant on the fan shaft/bearings or sleeve, so in effect you have none and premature failure is almost certain. If you re lubricate the fan shaft/bearings or sleeve, then this would of course be ok.

[BuddhaNature]

Hi Harry,

Good point - it might. Good idea, lubricate the fan. How do I do that on the stock fan that comes from IBM?

[Harryc]

http://www.ocforums.com/showthread.php?t=437021

[spuddog]

I think it is great when someone goes waaaay outside the box and fixs a problem. maybe it won't last forever, but it wasn't working anyway. Good job

Scott

[BuddhaNature]

Thanks very much for the link, Harry. Excellent stuff. Sort that out in a couple of days time though I'll get some alcohol before I do. Thanks again.