Diagnose my computer: load-related shutdowns

[jbsegal]

So my laptop - 4+ year old 15" MacBook Pro - has taken to shutting down under high memory/high cpu conditions.

Nothing is logged, it just powers off.

I've run Memtest on it for a number of cycles and it comes up clean.

I figure that either:
A) Memtest is wrong and there's a bad memory spot that only gets hit occasionally
B) There are thermal problems.
C) I have no other ideas.

Any opinions from the crowd?
(I'd love a resounding "It's the memory!" as getting to the heatsink to re-thermal-paste this is a challenge", but I'm not expecting it.)

Comments

[dalesql.livejournal.com]

Fits the classic overheating symptoms. Check your fan and all the air paths and air filters if any. A nice heavy layer of dust clogging things up will do this exact thing. Secondary check, is the fan actually going to full speed? If the fan is variable speed, check the temperature sensing to see if it is going full speed when hot.

[chienne-folle.livejournal.com]

My computer started shutting itself down every once in awhile, and it turned out to the the power supply.

[pir.livejournal.com]

Open it up and vacuum out dust, etc?

[c1.livejournal.com]

Taking apart an iBook (yes, not a MacBook!) isn't that bad-- a metric buttload of screws in places more obscure than Timbuktu, but certainly doable in about ten minutes. I can't imagine things inside Apple's Reality Distortion Field have gotten so extreme that current Apple laptops won't yield the goods to a similar effort.
Once open, following everyone's comments about blowing out/vacuuming dust should be productive.

[jbsegal.livejournal.com]

There's a reason I included a link to the teardown to get to the heatsink.

Replacing the hard drive wasn't too bad. This is a complete deconstruction.

[jbsegal.livejournal.com]

(Near complete. Just have to get the display cables out of the way.)

[dr-memory.livejournal.com]

Sadly, I'm afraid that sounds exactly like a thermal problem. Does the fan fire up?

[dr-memory.livejournal.com]

BTW, Apple's Genius Bar prices for service on old laptops are actually shockingly reasonable. They replaced the mobo and dvd drive on my old 12" Powerbook for (IIRC) $300 in 2008.

[en-ki.livejournal.com]

memtest sucks, compile Linux 10 times with "make -j"

[jbsegal.livejournal.com]

That page - from 2003 - links to http://www.memtest86.com/ which seems to have ceased development in '07. http://www.memtest.org/ is Memtest+ which is still being worked on. I'm going to assume they're getting better and better at what they do. :)

Now, I can't really speak to the current validity of the concept, but I really don't want to reset this system to build kernels on.

Thanks, though. :)

[librarian.livejournal.com]

Sounds thermal, you can try Cool Book to maybe underclock it

[dcseain.livejournal.com]

Sounds like overheating, as echoed above.

[inahandbasket]

Grr, got logged out. That anon comment was me.
To repeat:
Memtest sucks, use Prime95.
http://mattgadient.com/2008/03/29/prime95-for-mac-os-x/
You can stress the CPU or the memory selectively, it'll let you figure out what's overheating.

That said, it's an old laptop so the thermal paste is probably toast. I'd be willing to bet money that tearing it down and re-pasting the CPU would fix it. That stuff has a limited useful life.

[jbsegal.livejournal.com]

What's fundamentally wrong with Memtest (or really, Memtest+, the still-under-development branch)?

And yeah, I've heard that about thermal paste, which is why I started with figuring out what the teardown looks like and THEN continued on to trying to avoid it. :/

[inahandbasket]

I was a bit general in my statement, let me overhaul. ;)

Memtest, while great for error checking RAM, sucks for diagnosing overheating issues, as you can't separately test the various bits and pieces. Prime95 (or similar software) will let you zero in on which component exactly is overheating, from RAM to CPU to northbridge, etc.

[eccentrific.livejournal.com]

I agree with it sounds like overheating. I wouldn't start with the thermal paste though. Clean out all the fans, make sure they're spinning at speed (last time I had overheating problems, it turned out to just be cat hair in the vents...), clean out the vents, etc. Figure out what feels hot. It might be an individual component shorting and overheating, not the whole system. If all else fails *then* maybe try the thermal paste.