Bumping Early 2008 Mac Pro causes it to drop out.

[Mr. Paul]

Even the slightest bump on my Mac Pro (early 2008, 8 cores, Xeon 2.8GHz) causes it to:

- screen goes black
- All fans wind up to max
- system becomes complete unresponsive, including to remote access (no ping, etc)

Restarting generally makes everything ok (but you have to hold power button for 5s or so to get it to power off), though this morning it failed to start a couple times too; just went straight to fan madness.

This actually happens without bumping it as well on and off, but bumping it is pretty reliable. I've
- re-seated all memory. It kept 'forgetting' about some of the memory in Bank B, so I pulled all of it (down to 8GB now, dammit), it no longer forgets any of the installed memory but the problems continue and are getting worse.
- re-seated the video card
- Blew all the dust out

Any ideas?

 

[djjclark]

By bump do you mean a physical hit or more of an ESD? Basic things would be to make sure all parts are in firmly not just memory and all cables especially your external power cord. If ESD you might try a different outlet/cord to see if there is a grounding issue.

 

[Mr. Paul]

By bump do you mean a physical hit or more of an ESD?

Physical, though as I note if you just leave the machine it will fail eventually without a bump; recently, probably about 50% of the time overnight. This morning it was getting hard to use as it was failing very often.

All external cables are well seated. Maybe I'll take a run at re-seating internally again, but the last attempt was pretty thorough and didn't seem to help.

 

[macthefork]

I saw the link to this thread over in the iMac forum.

Sounds like you may have a graphics card failing. Black screen and Fans reving up could be signs of this. If there's a disconnect in the Graphics card, it also can cause the unresponsiveness of the Logic board.

Maybe try baking it. Yep, that's right... Google "Bake Graphic Card" for discussions of this and how to... This apparently works by allowing loose (cold) solder joints to melt and resolder themselves. If the Graphics card is what's failing, you've got nothing to lose. And many times, it fixes the problem.

 

[Tesselator]

Get your vacuum cleaner (and air compressor if you have one) and start breaking it down. Take everything but the CPUs out. Everything - PSU, HDDs, RAM, PCI Cards, fan assembly cages, the works! Blow out or suck out the dust as you go. If you don't have an air compressor it's worth borrowing one from a friend. Do all this in your garage where there is plenty of room to work and the dust won't kill your pets.

As you reassemble it wipe everything down with a very damp rag. I like using two rags. One rung out from a bucket of 5% to 10% kitchen bleach and a tiny squirt of dish soap. And the second final wipe with 30% to 40% rubbing alcohol. Hit all the ports too - YAY for cotton swabs and tooth brushes.

After reassembly turn a strong room fan on it for a few hours to make certain everything is dry. Turn it on and give that a try.

If the outages continue buy some thermal cream and some acetone. Remove your CPUs as per some on-line tutorials specific to your system. Use the same cleaning solutions to wipe down the CPU connection pads with a final swipe of a lightly acetoned rag or lens cleaner tish. Clean off the old thermal cream and finally use the acetone to make the CPU thermal spreader surface perfect (do not touch after that step). Then same for the heat-sink surface. Do the thermal cream thing (lightly, evenly, no fingers!) and replace the CPUs and the heat-sinks.

The first round of cleaning should probably either fix the problem or allow you to uncover the specific causes of the trouble. If neither of these things work then you need to have the machine repaired professionally. If you're not familiar with diagnostic procedures and/or have the equipment on hand to carry it out, you'll likely spend more replace-by-guessing which part needs replacing than just having the shop do it (probably).


PS: People should be dong that first round bit to their MacPros every year or so anyway...

 

[Mr. Paul]

I've done one moderately thorough clean (using an air compressor and vacuum) already, and re-seated everything easily accessible. It seemed to work I thought I had it licked, but it came back.

Then I noted the memory problem (forgetting memory); I removed the unrecognized sticks, though it was licked, but nope. Hard to tell if disturbing the case made it go away for a day or three, or if it was just luck; it is intermittent after all.

The memory thing made me assume it wasn't the video card, but on second thought, there is no reason why both couldn't be a problem but the current issue (blanked out, fans at max, hard stop) could be the video card and the memory was a side show (there were a few other much more intermittent problems). I actually have a spare video card (lower powered) and could swap for a while and see. I also have a hot-air reflow set-up, so if it is the card I could reflow the main problem chips, but don't want to risk destruction of the card unless I am sure it is the offending piece of gear.

I think I will strip it down again, even more thoroughly, use contact cleaner on everything and wipe it all down more thoroughly this time. I banged the video card in more thoroughly last night (while it was powered down) and it hasn't returned, but OTOH I've been there a few times....

Thanks for the suggestions so far!

----------

Oh, as a follow on, the memory (in case it gives anyone hints). I had 16 GB. It started reporting I had 12. I removed the 2 unrecognized sticks. A few days later, it started reporting I had 8. Unrecognized sticks were in B, same as the first sticks, so I pulled the entire daughter card. It seems to be holding steady at 8 now. Memory is third party, but has been running smoothly for 5 years now.

 

[Tesselator]

I dunno if you can keep your machine up long enough but the fastest way to actually test for bad memory beyond Apple's initial check, is to download Rember, duplicate it the number of times you have CPUs (one cope per core), open one and enter RAM/Cores (12GB ÷ 8?), set that much minus about 25MB in the setting window, turn OFF "Continue on error" and set "Quit All Apps" and "Quit Finder" in the preferences, quit and then open all 8 instances, and then just start them off. If it appears to be chugging too slowly or is one Rember instance seems almost stuck, stop (don't quit) that one and give it less RAM to do - a little at a time till it takes off normally.

I gotta say that actually sounds more like a faulty riser board than bad memory tho. (If indeed it's not a CPU seating problem - or a GPU problem as MacTheFork suggests).


There's also a CPU Test here: http://www.coolbook.se/CPUTest.html which is just a GUI for the processor's own full self test. I think this is never performed automatically by anything or any procedure without sending it the specific command manually.