Really weird PC hardware issue

[drachillix]

I have a machine in my shop with an epic stumper of a problem.

Dell desktop, came in with a cooked power supply.

We replaced PSU, fires up fine, only one small problem.

It only works laying on its side, if you try to tip it up into a proper vertical tower orientation it shuts off at about 15 degrees from vertical. Once in a vertical position it steadfastly refuses to powerup at all. I have pulled it all apart and reassembled, blew everything out with an air comressor, tried wiggling every single component I could get my hands on on the board/case/cards to see if its something loose to no avail.

Tried a different PSU as well, no difference.

[beowulff]

Check for loose RFI shielding, especially on any CD/DVD drives.

[Mangetout]

Obviously the PSU isn't the culprit (but you realised that already, I think)

Is there any chance there's a loose screw lodged between the motherboard and the case?

How about a loose connector on one of the jumpers that drives the case-front power switches or something.

When you tip it and it gets to the shutoff angle, can you see anything move?

Are there any microswitches that detect the case being opened etc?

[Mangetout]

Is the processor actually clamped into its socket with that little lever?

[Quartz]

Sounds like you've got a short somewhere? Which model Dell is this? What happens if you don't have the PSU screwed in but loose? Also, have you tried removing and reseating the motherboard? Is the case bent in any way?

[drachillix]

Originally Posted by Quartz

Sounds like you've got a short somewhere?

This is my assumption, been through all the suggestions above, pulled it totally apart and blew everything out to make sure there was nothing like a metal shaving or screw trapped in there and tipping against the board. I managed to get it to boot twice in an upright position but it will not stay on for more than a few second that way. No apparent damage or bending.

Its really weird because it seems to be truly a positional problem but there is noting I can find loose or bumping/dangling that might be shorting.

Its a dimension 8250 (tragic to see it go if its hopeless, its got a gig of RDRAM in it.

[Jurph]

When I saw that drachillix had a hardware problem I knew this would be a good one!

I would try sliding a thin piece of cardboard (like a manila folder) between the back of the mobo and the case, so that you can positively rule out case shorts; ever since my first mobo-to-case short I've done this to every machine I build. Saves a ton of pain and suffering. You could also take the case out of the equation and build the system on a tabletop with the case sitting nearby to provide power buttons. By adding components to the case one at a time (or rotating the components in place) you can check each component's sensitivity to rotation.

If the CPU heat sink is heavy, consider lashing it into place with zip ties -- it could be pulling one of the CPU pins juuuust loose enough to cause headaches. Test this by building the system without the case, and rotating the motherboard by hand. Your note about "no more than a few seconds" makes me think it's something like a pin somewhere being pulled into intermittent contact by gravity.

[Quartz]

If it were a heatsink issue, I would expect the PC to power up and scream loudly then power off. If the problem isn't a short, I wonder if there could be a loose capacitor?

[Suda]

I'd agree with the heatsink diagnosis. I fixed a PC with almost the exact same symptoms as you describe, drove me crazy. I found by accident that if I unscrewed one side of the heatsink just a little bit it'd work fine, but when I tightened it up it just wouldn't turn on. Can't explain why at all.

[drachillix]

Originally Posted by Jurph

If the CPU heat sink is heavy, consider lashing it into place with zip ties -- it could be pulling one of the CPU pins juuuust loose enough to cause headaches. Test this by building the system without the case, and rotating the motherboard by hand. Your note about "no more than a few seconds" makes me think it's something like a pin somewhere being pulled into intermittent contact by gravity.

HS is rock solid and clipped in so no loosening it without doing some creative bending of HS clips. I took it apart again and laid a sheet of that nifty anti static foam from motherboard boxes in there before reassembling. Same deal, problem must be on the mobo itself. Oh well, customer authorized it for a rebuild so its basically getting an upgrade and I'm going to build it in a new case to make sure. TY for the suggestions.

[KP]

Um ... I'm sure you know that antistatic foam is *conductive*. That's what makes it antistatic. Sure, not as conductive as metal, but putting random parasitic loads and low-medium impedance shorts all across the back of the motherboard would absolutely not my first choice in a debugging technique (or anything else).

I wouldn't have guessed it would start at all. I guess I learn something new every day.

You say it came in with a cooked power supply. While PSs can cook themselves over time, some oddball positional short could, too. However...

I'm wondering if it is strictly positional. For example some PCs have an mobo light sensor to protect against tampering (activated as a BIOS option). Tilting may expose it directly to some ambient light source. I saw behavior like that while repairing VCRs, due to the IR LEDs/phototransistors commonly used to detect (e.g.) tape and roller position. Did you try it with the case cover entirely closed? Entirely open?

Did you check that the Li coin cell was clipped in properly? BIOS batteries can cause funky symptoms, esp. during boot, and some holders are capable of partially clipping. I'd imagine you covered that when you reseated everything, but I had to ask.

I saw something like this from a free screw inside in a CD drive I fixed, but I don't recall the exact symptoms. I'm sure you tried it stripped down, though.

I once had to remove some grit (glass? rock?) from the bottom of a RAM slot with a dental pick (not just one particle, either) to get a computer reliable, but I know you tried reseating, but I didn't notice in several reseatings before it finally felt wrong . I don't know how it got there, but it wouldn't blow out with compressed air. (I promptly cleaned my bench)

I can see PCB microfractures doing this under the weight of some heatsinks, but that's a mobo problem requiring replacement. I can see a broken wire in the mobo wiring harness doing it to, but I've never seen that actually happen.

Sorry, just some oddball ideas for an oddball problem.

[Running with Scissors]

I'm looking at a Dell motherboard now (it's a bit newer, but the construction is probably similar) and I notice that there's some space between the top of the board and the bottom of the PCI and AGP connectors. Since a misaligned PCI card can prevent a PC from starting by shorting two pins, possibly a shard of metal got underneath and is doing the same thing.

[drachillix]

Originally Posted by KP

Um ... I'm sure you know that antistatic foam is *conductive*. That's what makes it antistatic. Sure, not as conductive as metal, but putting random parasitic loads and low-medium impedance shorts all across the back of the motherboard would absolutely not my first choice in a debugging technique (or anything else).

I knew that....funny thing was, it didn't make a difference.