What have you done with a PowerPC today?

[mac57mac57]

Today was "G5 Quad Day". I managed to acquire two additional G5 Quad CPU cards (not an easy task these days!) and used the first one to upgrade my Air Quad back to full Quad status and the second one to do the same to my LCS Quad.

Air Quad (Air Cooled Quad): Prior to today, only one CPU card was in use because the other was damaged and I had disabled it via an Open Firmware command. Today I installed one of the two "new" CPU cards into it and I am happy to report that the operation was a success. The second card even idles a degree or two cooler (38C to 39C) than the first one, which is a good result.

LCS Quad (LCS Cooled Quad): Prior to today, only one CPU card was physically installed, because the other one was quite dead and the machine would not boot with it physically present in the box. Today I installed the second of the two "new" CPU cards into it and this operation was initially a success. Like the Air Quad result, the second card even idled cooler than its running mate (36C to 39C) - a good result. Regrettably, an old bug-a-boo raised its head shortly thereafter: the cooling loop suddenly sprang a nasty leak. I noticed CPU B shooting up into the 90C area, followed by thermal shutdown. I found and repaired the leak and am letting the silicon sealant used for the repair to set overnight before firing the Quad up again. This episode simply reminds me that Air Cooling is far superior to LCS Cooling! Air coolers don't spring leaks!

So... I'll call it one and half out of two, or a 75% success rate for the day.

I am responsible for the new LCS leak of course - after discovering an air gap in the cooling loop while installing the second CPU card, I added more cooling fluid to the loop via needle insertion. I sealed the two holes (one to insert the coolant, the other to let air escape as the coolant took its place) as well as I knew how to, and then resealed them again half an hour later. Nonetheless, the new leak was at the exact spot where I added the coolant. Clearly, one has to let the silicon sealant set for more than one hour!

The loss of coolant does not appear to have been significant, so I will try it out again tomorrow after the sealant has set overnight and see if it runs. If not, I will have to tear it down again, refill, reseal and let it sit for another day. No worries either way however. This is a set back, not a fatal blow. The LCS Quad will ride again!

 

[mac57mac57]

I am happy to report that "Quad Day" was more successful than I thought. After my repair of the leak cured overnight, the machine booted right up and ran nicely. So, I'll call it 2 out of 2, or 100% success...

...except for one curiosity. The newly added CPU was CPU B. Core B1 idles at about 38C, but Core B0 idles at about 48C, a full 10C hotter, even though both cores are on the same physical die. Also, when I start up any "heavy" application (Aquafox, Photoshop, etc.), Core B0 spikes into the 80C range, while B1 remains in the 50C range. I have even had the machine put itself to sleep due to CPU B overheating - no checkstop, just sleep.

It is very curious that one of the two cores on a single physical die has such different thermal performance than its adjacent peer.

Since one of the two CPU B cores is nicely cooled, I have to conclude that the cooling loop itself is doing its job. I am wondering therefore if uneven thermal paste (part of the chip less covered than the rest) might be the culprit? Or perhaps a blockage in one part of the micro channels of the CPU B cooling pad?

Other thoughts?

 

[ervus]

I would guess that the copper block is not making flat/even contact with the silicon. Like it's at an angle or a small spec of dust/debris is lifting one edge. Copper conducts heat better than almost anything, so you want the silicon flat and tight against the copper. If you remove a heatsink and look at the residue of thermal compound, you can tell how good the contact is by the way it was squeezed out of the joint.

 

[barracuda156]

Fixed a modern Cocoa terminal with true color support:

It still relies on X11 for settings (apparently for any macOS, not just legacy ones), but terminal emulator itself uses native backend.

 

[mac57mac57]

Thanks @ervus, much appreciated. I have been thinking along the lines of inadequate pasting of the CPU, but you take that one step further... simply an uneven contact with the copper. I will have to check that out. Thanks again. Of course that does mean extracting the LCS from the Quad case and then disconnecting CPU B, ultimately necessitating a repaste of it as it is all put back together.

BTW, I love your avatar... the iconic "Think Different" phrasing, coupled with the original Macintosh logo colors superimposed on a Windows logo... brilliant, just brilliant!