Computer diagnosis required

[SCRawl]

I'm working on a friend's computer, and without just swapping out parts I'm not sure where to go from here.

It's a P4 small form factor desktop running WinXP. (Specifically, a ThinkCentre S51 (8173-KUB) which was purchased off-lease a year or so ago.) I will provide the symptoms.

Sometimes (but not often) when it boots up it gives some beeps, and my searches have revealed that this beep pattern (1-3-3-1) means: "Autosize DRAM". I ran MemTest x86 through nine passes, and it didn't find a memory error, so I'm thinking that memory isn't an issue. I made certain that it was seated properly, just in case.

I notice that sometimes the fans run at a very high rate of speed, so I thought that it might be overheating. SpeedFan reveals that it has a normal temperature around 63C, but gets up to 70C. (Normal fan speeds are around 2000-2300 RPM, but I've seen them get up to 3500 RPM.)

Sometimes the thing just quits unexpectedly.

Sometimes it doesn't even go to POST, and instead just sits there.

My next thought was to get a new power supply, though finding one to fit the form factor might be difficult.

Any thoughts?

[XaLEv]

Sounds like the problems I had with my old Athlon Thunderbird machine as it was dying. Look at the capacitors on the motherboard, and the ones in the power supply if you can see them, and see if any are bulged or leaking.

[Dominus Atheos]

A failing power supply isn't going to give those symptoms. If you've checked the memory, it's almost certainly a motherboard issue.

[SCRawl]

I can't see any obviously defective capacitors on the motherboard.

I took the heatsink off of the processor for a look, and it's possible that it isn't making good thermal contact. If that's the case, then I'm thinking that the processor has been damaged already. There are times when the thing can run just fine for hours, but other times it'll quit after a few minutes.

[Tolya]

Swap RAM to some other single (!) chip and see if it helps. All the symptoms you have described are quite typical for a faulty RAM chip, or the mobo.

RAM memtests of all sorts are, quite honestly, buggered liars. They are good for one thing: if you find an error then you get a confirmation that the chip is dead. Even if you test memory to be fine, you may still get errors/crashes during normal operations - because typically memtest runs a small amount of tests on a single memory location and if the system is working over a given time, that workload is much bigger than during a test. It may be that RAM does something fishy 1 out of 1000 operations - and you might have been lucky

Being mostly a homegrown techie, the above explanation may seem amateurish to somoene with professional education, but I've seen so many computers with RAM chips stealthily gangraping systems that for me they have become the first culprit in the "I have no idea what's happening" situations.

Also, are there two RAM chips or one? At least, RAM chips should have the same timing. At best, they should be their identical twins, i.e. same size, same timing, same manufacturer even. RAM chips with different specs tend to give exciting results. If there are two, take out one, leave the other at DRAM0 and see if the problems persist.

I wouldn't rule out faulty capacitors or some such (easy to spot, there is a bulge at the top), but Autosize DRAM error is motherboard failing to recognize the RAM chips to be valid and compatible, so I would start there.

[SCRawl]

The thing is that it's such an intermittent failure. I've had the system for a couple of days, and out of maybe 50 boot cycles it's given me the beep code only twice. It will almost always boot into Windows, and (almost always) after a short-medium time, it appears to overheat and shut itself down. I say "appears" because the temperature monitor doesn't exactly go sky-high -- only as high as 70 C that I've observed -- but the fans start speeding up (to about 4000 rpm, the software claims) and then it shuts down.

I don't think that I have any RAM around here of that vintage to swap it out, but I suppose I can just try it with one DIMM at a time, since each one is 512MB. (They both have the same specs, right down to the manufacturer, though.)

[SCRawl]

I just tried it again with the one DIMM, and the symptoms are the same as before. I did observe one thing, though: it may well be overheating without reporting a rise in temperature, because the system seems to lock up just as the overheating starts. I can't even move the cursor around the desktop with the mouse.

So I'll try improving thermal contact with the heatsink. And the other DIMM on its own.

[SCRawl]

Well, shit.

I watched some YouTube videos about how to clean the CPU/heatsink and then reapply the paste -- it isn't something I've done before, as most of my hardware experience has come from older hardware -- and then performed the operations myself. After putting it all together, it won't even go to POST. Instead, it pauses for a moment and then turns the fans on full blast, but that's as far as it goes.

I figure that I must have damaged the CPU or the socket when I removed the old paste. That stuff doesn't come off easily, and required a little (gentle) pressure.

On the bright side, it seems to be distributing heat a little better to the heatsink. The really fucking big down side is that it's now a boat anchor.

[Zaune]

Don't berate yourself too much, it's unlikely that anything you did could have killed it unless the board or the CPU were damaged already. A loose contact in the socket somewhere might well explain a lot of the symptoms you describe, actually.

Although come to think of it, but you did double-check that the connector between the motherboard and the power supply hadn't come adrift when you tried booting it back up, right?

[someone_else]

Did you use non-conductive paste?

It should perform slightly worse than conductive paste (you care about this only on gaiming rigs) but has the bright side of not shorting stuff if you aren't an artist in its use.

[Torben]

At this point, I would disconnect everything from the board except power and fans. I would even remove the RAM, and if you have a large anti-static bag, take the board out of the computer and place it on that. This will ensure that if there is an issue it lies with either the processor or the board itself, removing the possibility of any issue with shorts, bad peripherals, bad memory, etc. If the board POSTs and gives an error, then at least you know the board can turn on - does not necessarily mean it is good, but it is at least not definitely bad. If the board will not POST with these minimal connections, the issue almost definitely lies within the processor or board, and of the two, most likely the board.

I second the notion that more than likely your removal of the existing paste and application of new did not harm anything, unless you applied so much that it ran over the sides and possibly got mixed in with the pins...but, if you followed directions and only applied a thin film, then the whole situation was most likely doomed from the start and all you did was confirm the cause of death.

If the board does POST, then you can being reconnecting the peripherals, one at a time, powering on after each. If you receive a failure after one piece is connected, remove that piece and try again. If POST succeeds, then you may have your culprit, but continue the process until everything is connected.

Good luck! Let us know how it goes.

[SCRawl]

Thanks for the tips, folks, but I'm pretty sure that the patient has died.

The MB+processor (with no RAM or peripherals) just spins the fans, and won't POST. I know that the power supply is connected because the heatsink starts heating up. I looked very carefully at all the contacts, and nothing appears to have been contaminated. The paste is non-capacitive (Arctic MX-4, if anyone cares) and is only on the interface between the CPU housing and the heatsink block, and it's pretty thin.

The computer belongs to a friend's parents, and while the friend is at least as capable of computer maintenance as I am he lives too far away for a service call. Fortunately this is an old, good friend, and he knows that I probably didn't cause the death of it (and besides, it's a cheap off-lease computer). Indeed, the state of the paste -- it looked as though it had been baked on -- suggests that it died of excessive heat. The timing of the expiry makes me think that I must have done something, but death was probably inevitable.

What would be ideal would be to find the same model of computer (or something very similar) and just pop in the hard drive. That would be a really easy fix, I think, and it might be possible to find one. More likely, though, is moving the peripherals to another computer and returning that to its user.

So, yeah, thanks again for the advice.

[Enigma]

You can get a replacement for at least $64, I believe.