Mac G5 Heatsink Upgrade Question

[Harbinger]

I got a PowerMac G5 dual 1.8 GHz PowerPC machine at a great price, so I thought I might play around with it a bit (I'm a PC guy). After acquiring the machine I read a large number of online complaints about these units failing due to 'thermal runaway.' It seems that Apple and IBM had some issues with these, to put it mildly.

I decided to investigate. I loaded a program called Temperature Monitor and observed the idling temps, about 41 degrees C on each CPU die. Then I loaded and ran a program called CPUTest to stress the CPUs and observe the temps. Both die temps, over about 4-5 minutes rose to about 58 degrees C. Next I pulled the CPU/heatsink modules out and examined them. The CPUs are mounted to small PC Boards with some support components, and strapped to the bottom of this huge passive heatsink. I pulled one assembly apart, cleaned off the existing thermal paste, carefully applied a 'silver' heatsink paste, reassembled the CPU and the machine.

Now when I test, both CPUs seem to idle at about the same temperature, 41 degrees C. When I start CPU test, however, the modified CPU jumps 10 degrees C to around 52 degrees C almost instantly, then rises to 58 or so and stays there. The other, unmodified CPU will start climbing slowly over 4-5 minutes to about the same temperature.

Apple wants their service dealers to run a thermal calibration program on the machine when CPUs are changed or problems arise, and I have done so. I know that the silver paste will take some time to 'settle in', up to around 200 hours accorcding to one online report I'd read. What I'm wondering is whether or not this fast temperature rise behavior is "normal." Any input appreciated...

[billbartuska]

Generally, a fast CPU temperature rise indicates a ineffective heat transfer between the CPU and it's heat sink.
Just a guess, but the new thermal compound may not be making complete contact between the CPU/Heat sink (thus the fast temp rise) but be more effective at transferring the heat (and thus able to maintain the same maximum temp with less contact area). I'd check the CPU and heat sink for flatness.

[Harbinger]

Originally Posted by billbartuska

Generally, a fast CPU temperature rise indicates a ineffective heat transfer between the CPU and it's heat sink.
Just a guess, but the new thermal compound may not be making complete contact between the CPU/Heat sink (thus the fast temp rise) but be more effective at transferring the heat (and thus able to maintain the same maximum temp with less contact area). I'd check the CPU and heat sink for flatness.

Thanks for the pointer! I went out tonight and got a tube of real 'Arctic Silver 5' (which, oddly enough, isn't silver like the knockoff compund I was using but rather a gray color). I tore the assembly down again, cleaned off the silver paste, then used a razor blade to carefully spread a thin layer of the AC5 onto the CPU. I then reattached the board as recommended (with one screw on each corner of the CPU, a turn at a time, opposing corners) so all was tight again.

Now my temp on that CPU is only 5 degrees C higher under moderate load than the unmodified CPU, not 15 or so. They're both extremely close at idle.

I don't think I can get any tighter/closer match or bond between the CPU and the 'sink. I'm going to put the machine through heating/cooling cycles and give the paste a chance to settle in now. I'll let you know what happens.

[ben333]

Think you could take a few pictures? All you really need is a good mount and just a small amount of thermal paste but it would be interesting to see your CPU/HS setup.

[billbartuska]

Originally Posted by Harbinger

...which, oddly enough, isn't silver
... used a razor blade to carefully spread a thin layer of the AC5 onto the CPU.

AS5, like most TIM, is a suspension of tiny silver particles in some "goo" that squeezes out and leaves mostly just the silver particles between the HS/block and CPU.

AS5 instructions

Yes, pics.