Schrödinger's MDD G4 - dead or alive?

[CJ_Miller]

Recent problems with my MDD inspired me to get my first TiBook as a portable media machine, but I still need to get my tower working again. It is so flaky now I almost with I could write it off as dead, but it was solid for years, and now that I am putting together a new system there's a lot of stuff I want off of there. This weekend I am trying to be methodical and try everything I can think of to get it running. I can't afford any more parts or tools now so I need to work with what's here.

It was a 166 MHz FW400 logic board with a dual 1.42GHz processor card and copper heatsink installed from a partial FW800. The symptoms? It kernel panics constantly in 10.4.11, and just hangs in 9.2.2. There are serious bus issues here. I did try the Apple Diagnostic Disc which I have here, but it is of no use, it claims it is for a different model G4. Here's what I have tried so far, if your MDD be buggin' feel free to use the checklist to get it sorted out:

- ATA 100 bus - changed jumper and booted from X drive only

- ATA 100 bus - " from 9 drive only

- replaced GeForce Ti 4600 card with Radeon 9000, and then another Radeon 9000

- cleaned AGP, PCI, and RAM slots

- removed all PCI cards

- tried different RAM combinations

- removed all USB and tried different keyboard and mouse

 

[CJ_Miller]

TechTool Deluxe:

I've got a disc which was given to me with one of my partial MDDs, maybe 2.7.3?

- booted TTD for OS 9 drive (x removed)

It scanned my hardware and RAM before it hung while scanning volume structures.

- booted TTD for OS X drive (9 removed)

It restarted my mac a few times and panicked with a memory error: unaligned frame address.

So typical of tonights headaches, two versions of TechTool tell me two different problems.

- Also I retried this with my two ATA 66 drives disconnected, just to narrow it down a bit more.

Anybody have any idea why my MDD decided to die when I wasn't using it? It ran practically every day for about five years! I switched it off only if I was (rarely) going to be away for a few days, or if there was a bad storm. And once I finally leave it off for a few months, it dies. This is not the first time computers have been this way for me, they go from constant use to dying from apparent non-use. I am really curious to know why, if anybody has ideas! I guess I am going to search my house now for:

1. spare RAM (doubtful)

2. new ATA cable

3. other (untested) MDD CPU

 

[CJ_Miller]

Alright, I am tired of messing around with it today...

I have three sticks of compatible RAM, which at various times have run in this box. None of them work. One of them allows it to boot, and it promptly freezes. With the other two it chimes on but doesn't even start to boot. What are the chances of three good sticks all being ruined at the same time? I am running out of ideas. It sucks to waste my whole weekend working on this thing and not getting anywhere.

 

[CJ_Miller]

IT'S ALIVE!!!

I collected all of my MDD bits together tonight and set about methodically testing. First I tried combinations of RAM sticks - they can't all have gone bad! And unlike last time I messed around with this box, I had found my stash of spare MDD CPUs, of which I was told one was good, and the other fried. I seem to remember being disappointed that the good one was slower, which made it the single 1GHz card. I cleaned the old thermal compound off of my heatsink and G4, dusted a bit, and let the alcohol dry. When I reassembled with new compound, threw two 512MB sticks of PC2700, and it started just fine. Actually, the freezing of this thing had gotten so bad that instead of running it for a few mintes at a time, I couldn't even boot without a freeze or a panic. So this is the first time I have even seen it boot in months. And so far there have been no problems. I am too psyched.

My dual 1.42 GHz card is indeed dead. On the lookout for a deluxe dual 7455 upgrade. At least it should be running much, much cooler now.

 

[mac-man6]

So the problem was the CPU daughtercard? It seems this is a common issue with overheating.

I've experienced it viewtopic.php?f=13&t=13430

And other have too viewtopic.php?f=13&t=11749

I've you have an MDD you might want to change the paste before it's too late. Or if you come across one make this could be an easy fix.

 

[CJ_Miller]

I am amazed that it didn't happen sooner! My MDD was regularly +150F, yet it didn't complain for five years. Meanwhile I had a feature creep situation with a project for recasing it. In fact, when it died, I was in the middle of distilling hundreds of CDs into a few DVD-DL of essential utilities and programs so that installing and maintaining my systems would be easier, I only needed another week or two. Now my discs are back to being as disorganized as they ever were!

Keeping thermal compound on hand is a good idea. My old school Macs didn't really need it, this was my first box where cooling was a major issue. Also I needed some for repairs I friend asked me about. I got fooled by the error messages referring to memory, the problem was the CPUs use of it.

I will see how this single 1gig goes, I expect it will be remarkably cooler than my dual 1.42 was. I have a naked spare MDD case which I will start retooling as a mounting plate and port panel for my next faster model. Even just a roomier case will help, with a full-length PCI video card, full-length PCI audio card, four hard drives, two opticals - sure, it all fit, and basically worked - but in a full size server like tower this stuff will be much happier.

 

[johnklos]

Regardless of whether a machine appears fine, you should replace the foil between the CPUs and the heat sink the first time you take the heat sink off.

Heat sink compound is always important. The foil doesn't work in the long term because the spring clips are constantly putting tension on the CPU board and over time it eventually warps a little. It doesn't take much to make a gap between the CPU and the heat sink that the foil can't handle, but a pasty material will allow a little gap and still conduct the heat.

Remember, everyone, ALWAYS repaste your CPUs if they have original paste or foil on them, and recheck them every few years.

 

[Paralel]

I'm glad I did that to the CPU card I got for my MDD then. Damn thing seems to run hot enough as it is.

 

[H3NRY]

Similar experience with my Sawtooth running a 1.2GHz PLX CPU. In my case the DIP switches which set the buss multiplier evidently glitched and it went to 1.6GHz, which is way overclocked for a 1GHz 7455 (PowerLogix had already overclocked it by 200MHz). It overheated and the symptoms were the same - memory errors, buss errors, and just generally flaky. I'm surprised it worked at all at that speed, but it did work apparently for a week or so before it finally kicked the bucket. I had noticed the fan was running at high speed a lot, but didn't think to investigate until it started crashing. It's possible of course that the bad part is a cache RAM chip instead of the G4.

 

[johnklos]

Similar experience with my Sawtooth running a 1.2GHz PLX CPU. In my case the DIP switches which set the buss multiplier evidently glitched and it went to 1.6GHz, which is way overclocked for a 1GHz 7455 (PowerLogix had already overclocked it by 200MHz). It overheated and the symptoms were the same - memory errors, buss errors, and just generally flaky. I'm surprised it worked at all at that speed, but it did work apparently for a week or so before it finally kicked the bucket. I had noticed the fan was running at high speed a lot, but didn't think to investigate until it started crashing. It's possible of course that the bad part is a cache RAM chip instead of the G4.

That sounds like a 1.2 GHz CPU card configured for a 100 MHz bus but in a system with a 133 MHz bus (12 * 133 is 1.6 GHz). But you say it was in a Sawtooth, which has a 100 MHz bus... Strange...

 

[CJ_Miller]

It's possible of course that the bad part is a cache RAM chip instead of the G4.

ARG!!! Why can't I just let it die in peace?

It worked for about three days, and now it is dead again! I was setting it up for work this weekend and had no problems. Next time I tried to start it to actually do something useful and it froze. Now it hangs early in the boot process, about five seconds after the chime. At least it didn't wait to die until I had some work half-done on it, so it could have been worse.

I guess that swapping out the keyboard+mouse, video, RAM, PSU, drives, and CPU - that this leaves the board itself. So I broke down and ordered a "new" 166MHz FW400 logic board.