G5 Quad won't calibrate...& other issues

Question

Wow, I've been through the ringer attempting the seemingly straightforward task of combining the working parts (which includes most of them) of two parts/repair G5 Quads into one working Quad. As a consultant and troubleshooter, I expect computers to respond consistently in most cases. I forgot I was working on a G5...
Okay, where to start...? Well, the main thing I'm tackling right now is the joyous Thermal Calibration task. Trying ASD 2.6.3 on 2 separate Quad chassis and logic boards (and 1 CPU module), I am unable to get through the Calibration procedure.
Some questions and observations:
1. Thermal Calibration will sometimes give me a fan-related error, but usually just dies and locks up the computer entirely. In this case, the Overtemp LED (#2) comes on, along w/ the Checkstop LED (#7).
2. Despite this, the computer will boot up and seems to run fine initially. But then w/o any significant CPU load or heating, the fans and pump will gradually increase (we're talking over 5-10 minutes) in speed until my G5 is ready to taxi down the runway and take to the skies.
3. CPU B Core 2 seems to be the most temperature sensitive, occasionally hitting 30-40F higher than the others, but it seems to bounce around somewhat. While the others are in the 105-115F range at idle, B2 jumps around from 120-155F.
4. Would another version of ASD be able to complete the Thermal Calibration? Apple Service Source shows 2.6.3 as the correct version for the Late 2005 G5s. Does anyone have experience w/ calibrating using another ASD version?
5. I have tried so many combinations of CPU cards, number of CPUs, chassis/logic board, etc, that I am now at a loss. When I think I have it figured out, one combination seems like it should work and is working, something screwball will happen like the fans ramping up or a random lockup. I'm tempted to toss all of this out in the street. I like a challenge and project, but when it doesn't behave consistently, I start lamenting my time wasted w/o the satisfaction of hard-earned success.
6. Here is what my component testing has shown and how I've combined parts:
Quad #1: 2-pump LCS w/ the lower LCS loop failed
-lower CPU card possibly compromised from this, although it seemed to work fine in a single-card arrangement
-logic board, power supply, and upper CPU card tested fine
Quad #2: 1-pump LCS w/ bad upper CPU card
-logic board, PS, and lower CPU card tested fine
So I have made 1 supposedly functional CPU module using Quad #1's upper CPU card and Quad #2's lower CPU card and 1-pump LCS. I have tried this Frankensteined CPU module in both Quad chassis/MLB combinations, w/ similar results. Main difference I have noticed is that Quad #1 doesn't like Quad #2's LCS, causing the pump to run (or think it is running) at 3600RPM and error out on ASD pump testing.
Sorry for the novel, but I'd appreciate any thoughts, answers or comments to help restore my sanity. I had thought that a "last, greatest Power Mac" would be a cool system to have, but now I'm wondering if I shouldn't have just saved my time and dropped a bunch more money on a Mac Pro instead.
Thanks,
Fred
P.S. Sheesh...as I typed this, naturally the Quad snickered behind me and proceeded to idle along happily for 30 minutes, then properly put itself to sleep...and properly wake as well. WTH??

Answer 1

DaddyPaycheck wrote:
What leads you to believe you have two good logic boards?
Heheh....an excellent question indeed! At this point, I'm not too sure of anything anymore... But seriously, to answer your question, a few things make me think the logic boards might be ok:
1. The machines have come up and run w/o failure, overtemp, etc, with only 1 CPU card installed (making them essentially dual 2.5s). In fact, last night I ran a Compressor batch as a test w/ one of the "half Quad" configs serving as the cluster controller and processor, and it ran w/ maxed out CPU activity for about 26 minutes straight. This was after running for a couple of hours playing iTunes and surfing the Web. Temps on the installed CPU's 2 cores never reached above 150F while I was observing, and fan speeds did not ramp up into the stratosphere as before w/ both CPU cards.
2. I figure that, unless I purchased these machines from really nefarious people (and I don't think I did) who combined more than 1 failed item into a Quad chassis in order to get rid of it, the odds of having both a failed LCS or CPU and a failed MLB are very small (in the absence of corrosive residue, I'm assuming an LCS failure didn't cause an MLB failure). Since I've determined that one Quad has a failed LCS leading to overtemp, and the other a bad CPU card, I feel I most likely have proper logic boards.
3. I haven't yet been able to produce a logic board failure during ASD testing.
That last one is less significant to me, as I've seen RAM and CPU failures that were not caught by ASD. But all 3 of those items together make me think there's a high likelihood that my logic boards are good (although failing to get through thermal calibration may be a sign to the contrary, but they run okay). Not a guarantee, of course...
For documentation's sake, last night in Quad #2 (which I've been working from for the last bit), I pulled the CPU module, then removed Quad #1's CPU1 card from the lower location, replacing Quad #2's CPU2 card in the upper location in the process. This left the lower location empty, and halved the Quad to a Dual 2.5. When I did this I started over w/ thermal paste, removing all of what was there and applying new (I had not completely removed the existing before). As I said above this has worked well even doing a Compressor batch. My next step is to reinstall a lower CPU card, also after full thermal paste cleanup.
Thanks for reading and your thoughts. I welcome any additional suggestions and ideas.
Fred