[scratt]

I have a Core Duo 17" MBP with the ATIx1600 GPU. Here is my story..

Whilst I have swapped HDs and RAM and stuff in virtually every machine I have ever owned (and feel I could do that in my sleep) I have never gone as far as stripping out the logic board and so on..

Like a lot of those early MBP models mine has been exhibiting worsening symptoms related to what is most likely GPU overheating, or worse.. There is a rather large thread over at Apple about this, and they are, I guess, not seeing the same numbers as with the 8600 NVidia problem as they are not reacting to it at all. Or perhaps they don't want to because ATI won't underwrite it for them...

Anyways..

My symptoms include:

• Vertical line on Display Panel. (Strangely always in the same place and going away when the machine gets warm.)
  
  Rare initially, but increasingly frequent refusals to wake from sleep, or simply a seemingly comatose machine.
  
  Throughout the machines life I have had a lot of GPU crashes I simply could not fathom, which other people could not replicate when running my code.. But you tend to put that down to Sod's Law.
  
  Increasing very recently are artifacts when the GPU is under load, which last week could be made to happen anytime I used Expose to move more than three or 4 windows around.

The general consensus seems to be that the next stage after the last in my list is Game Over! And as much as archiving a machine that is relatively young is a shame I am not going to pay close to $1000 for a logic board, nor am I going to buy a second hand machine to replace it as most likely any MBP comparable to this is going to have the same possible issues coming up... Having said that I do need a second Intel Mac for my workflow...

As well as all that the machine has always been pretty flakey wrt to USB and even Bluetooth mice, and a lot of this seemed to get worse when the machine was hotter. That always seemed strange. The machine would get hot and then my mouse would start disconnecting and reconnecting. But you live with these things..

So yeah, I seem to have got a "Friday machine" which is now very much on it's last legs.

Typically my MBP CPU would idle at around 55C - 65C and very easily hit 80C - 90C, and under a long period of load has almost got to 100 Degrees Celsius on rare occasions. This has been pretty much it's temperature range since I got it just over a couple of years ago. For most of it's life I have run it with a small desktop fan next to it.
Again, I put up with the fact it was hot and put a certain amount of that down to the fact that I use it in 35+ Degrees heat anyway.

Overall even with these problems this MBP has done a good three years (almost) and more than paid for itself so I can't complain. But it seems a shame to write this off when it sits next to an 8 year old PB G4 17" which is still going strong!

With nothing to lose, as Apple are not interested in this issue, and I have no warranty anyway, I decided to have a look at the state of the cooling area and thermal paste.

Using one of the iFixit tear-downs (which are superbly accurate) I got the machine stripped down and the logic board out in about 30 minutes.

The thermal paste was really pretty funky, and there was a lot!! But having taken the machine apart and put it together now, I can understand how hard it is to really gauge that you've got it on right, so I guess the guys n gals in the factory go for the more is less approach. Not a good plan really though as it actually makes heat dissipation less efficient.

Anyway I cleaned it all off, and a fair amount of light loose tumbleweed like dust that had built up around the cooling pipes, seemingly helped into existence by the gobs of excess thermal paste oozing around everywhere.

I applied what I feel is a correct amount of thermal paste, to each of the three chips down there. (I was surprised there were 3 actually.) Shows what I know about the innards of the MBP. Anyway the paste was some stuff I picked up here from a specialist PC store for a couple of dollars. Gold, shiny, weird shit. I used it to fix the same problem in my 360 a while back.

I had a few fun things happen when putting it all back together..

Even though I had taped every screw to a piece of paper and labelled them, I still managed to have one screw slot missing a screw at one point.. Always happens!

The machine would not start up at all after putting it together again. I simply got either a comatose Mac with a dim lid latch light, or the urgent "I am screwed" lid latch light pulses.

So I guessed moving the logic board around had actually killed it finally.

Not wanting to give up totally I pulled all the ram out to see what happened. Perviously I had fixed zombie issues like this. It obviously would not start up, but it gave the right error code.

So I reset the power management unit and put one stick of RAM in. She booted fine.

Except the screen was borked completely. Basically it looked like I was running a 256 colour game on an old 4 colour PC.

I put that down to me having damaged, or not reconnected the display connector properly.
Booting into Single User mode confirmed that as the display was readable, except the screen had a perfect red background!

With it in that state I ran everything up and did some heavy load testing on the machine to see if anything I had done had done any good at all..

I was amazed!

The CPU temp was showing 45 Degrees C. Fans were at 999RPM.
I repeatedly ran every benchmarking app I have back to back: GeekBench, Cinebench. Super nasty maths profiling loops I have for SSE development. My own OpenGL Procedural Planet Sim, which will easily crank the machine up to 90 Degrees normally, and uses the GPU to the max as well as both cores...

The highest it got to was about 66 Degrees C, and the GPU was running super cool all the time.
The fan peaked at 2500RPM. (Previously my fans idle speed was 3500 RPM!)

Enthused I stripped it down again and figured out how to boot up with the keyboard section of the laptop off.
Poking around at the video cable with a spludger I maned to get the display to change a bit. So I persevered and found it was about a 10th of a mm out more on one end than the other. I gave it a nudge that I was a bit too scared to do earlier in the day and that did the trick...

Putting it all together one more time I have come to the conclusion that one unrelated problem is that one of the RAM slots has a bad connection of some sort, or one of my Apple supplied SIMMs is going bad, as the machine is rock solid with one SIMM in, but I have had a couple of issues getting it to start with two.

Swapping the two SIMMs over has sorted that for now.. So it may be dirt, wear.,, Who knows? But I have ordered some new RAM for it now anyway.. And with even one slot out of order it is still going to be usable.

Overall I had a machine that was destined for the store room last week that is now perfectly usable again.

For how long, who knows..

But based on what I have seen, the second that my Unibody 17" is out of warranty I'll be extending my usual HD replacement procedure into a CPU / GPU thermal paste service as well....

If you have a MBP that is out of warranty I would seriously urge you to either get someone to look at it it for you, or take the plunge yourself. The whole CPU / GPU heat / thermal paste problem seems to have affected a lot of manufacturers over the last couple of years. And my machine was certainly a perfect example of sloppy manufacture which could be easily avoided.

Just thought I'd share this in case it is of any help to anyone else.
Sorry for the long post, but it was a fun day today and I wanted to share!
A good thing to have on the tv while you are doing this all is something like the LeMans 24 Hour race.

btw If anyone is considering doing anything similar and has any questions I am more than happy to field them.