Finally recovered an old MDD G4

I bought a Dual G4 1.0GHz MDD from Free Geek Minneapolis for $10 before the PPC 2024 Challenge. It never powered on. I built my own ATX power cable adapter and I got it to power but never chime. I just purchased a dual 867MHz mobo/daughter from eBay and swapped it out. It booted fine. The original was a 167MHz motherboard, and the dual 867 is a 133MHz motherboard. I noticed that when I removed the heatsink on the 1.0GHz, there was some burnt residue and the PPC chips had some dark corners. So, just to play around... I found the tables.

I down graded the bus speed on the 1.0GHz daughter (red) to 133MHz, and set the clock to 933MHz (yellow) just to test it out on the 133MHz motherboard. Still no boot chime or video. So the dual 1.0GHz daughter is dead.

I upgraded the dual 867MHz daughter to 167MHz bus (blue) and slightly overclocked it to 917MHz (Green).


I still have to revive a couple of my PCI-SATA cards to put an SSD in here.

As it turns out, this machine had a bad PSU and I had utilized a modified PC power supply.

I was going to sell this unit, and I fired it up again to make sure everything was fine, but it was not.

Shortly after boot, it would freeze up. And I could not get it to boot/chime properly. I recently picked up an actual MDD psu as salvage from the University of MN ReUse center that was a bit hacked up.

I tested all the voltages and they seemed within range, so I opened it up to blow out the dust... but there wasn't any!

I also grabbed a spare P1 24pin power connector from UofMN and proceeded to move all the good pins to that connector.

Then I repaired all the cut wires. And plugged them into P1.

The machine powers on, but still no boot/chime/video. Something else is wrong. I revisited the CPU clock resistors and took it down to 833MHz. Still nothing. I noticed the thermal paste was not fully flattened (I forgot to photograph it), so I cleaned it off and added new thermal compound. Further testing and still no success. It has to be something else. When I removed the heat sink, the new thermal paste was not flattened all the way either. Hmmmm... Closer inspection of the CPU daughter card shows some warping. The heat sink is obviously NOT making full contact with the CPUs. So, they could be burned out at this point. The previous daughter card had burned CPUs.

Photo above shows a slight bow up on the top edge (when case is closed) of the card. Photo below shows slight bow down from the rear edge (when case is closed).

Closer inspection of the CPUs themselves, shows they are pre-production 933MHz?

Options? Thermal pad will likely not push enough heat to the sink. Shims? Yes, but what thickness with this warping? I am going to use an old school technique from the automotive industry. Plastigage. O'Reilly Auto Parts only had the green in stock, so I ordered the other colors and they should be in tomorrow. I will give it a go with the green now.

Placing a small piece of it diagonally on the CPUs so I can determine where the actual contact is being made. To make the CPUs right reading, I am viewing the card from the rear of the case at the bottom of the photo, and the front of the case towards the top.

Install heat sink. Remove heat sink. And that tells us the the left (top) CPU is not making contact at all, and the right (lower) CPU is making partial contact. Both green strings should be smashed flat and evenly.


Tomorrow I will retest with the other thicknesses of Plastigage and see how large the gap actually is.