Computer keeps turning off - what needs to be replaced?
August 11, 2011 6:39 PM   Subscribe

Please help me identify which part in my desktop computer needs replacing! It keeps shutting itself off.

A few days ago, my WinXP computer started shutting itself down. I didn't see anything in the event logs that clearly indicated a culprit.

Today, I came home and my computer was off. I keep trying to reboot it, and it will stay on for no more than a minute. I have yet to have it stay on long enough to even see the WinXP boot screen.

I suspected an overheating CPU, so I went into BIOS, monitored CPU temp at ~ 42C. The computer was in BIOS for about a minute and then shut itself down, so clearly some piece of hardware is failing, right?

I just don't know where to go from here!

Is this indicative of a motherboard failure? I see no bulging capacitors.

Power supply?


What tests can I do to identify the problematic hardware?

I've made no hardware changes to this computer in over a year.
posted by lilgoyl to Computers & Internet (17 answers total) 1 user marked this as a favorite
The first thing I would do is connect a new power supply up to it and see if that solves the problem. There are power supply testers out there for 20 or 30 bucks .... but that's about the same price for a cheapo power supply as well.
posted by Poet_Lariat at 6:42 PM on August 11, 2011

power supply, this has happened to me a few times. if you pay $40-50 for one it will never break.
posted by miyabo at 6:45 PM on August 11, 2011

When this was happening to me on my Franken-puter, the culprit was a bad fan on my vid card that led to the whole thing overheating. I'd check your cooling system/heat sinks before changing out the power supply. Even if the temp monitor isn't showing a high temperature, it's worth a look.
posted by bluejayway at 6:49 PM on August 11, 2011

I didn't see anything in the event logs that clearly indicated a culprit.

This is a thermal problem. Open the box outside, blow out the dust with some compressed air and make sure all the fans (cpu, gpu, case, power supply, etc) are spinning when you turn it on. One of them isn't, and this overheating is what's causing the failure.
posted by mhoye at 6:58 PM on August 11, 2011 [1 favorite]

This is my power supply:

Purchased a year or two ago.
posted by lilgoyl at 7:01 PM on August 11, 2011

First thing I'd accuse is a failing CPU fan or dust-filled heatsink -- modern computers are good about shutting themselves off when the CPU overheats, and then after it cools down it'll turn on again like nothing happened; it won't always even show up in the logs, because it's at a motherboard-level, and Windows logging doesn't have enough time to log it before things shut off.
posted by AzraelBrown at 7:09 PM on August 11, 2011

If it was an overheating CPU, wouldn't it go above 42 degrees C when I'm monitoring it in BIOS?

I just witnessed it shut off when the CPU was at 42C (according to BIOS hardware monitor).
posted by lilgoyl at 7:23 PM on August 11, 2011

If it was an overheating CPU, wouldn't it go above 42 degrees C when I'm monitoring it in BIOS?

Most likely you are right. It's always possible that the BIOS temp update interval is long enough that you are not seeing the temp rise before the machine shuts off. But I think the most likely culprit is the power supply as well. I see you bought a new one 2 years ago - that doesn't mean that it isn't broken now :)

Here's a quick diagnostic you can do to possibly see if a temp increase is involved. Open the machine .Get a can of compressed air. Turn the can upside town and do not use the tiny tube that it comes with. Power up the machine and while it is running give the CPU several blasts of the coolant that will be coming out of the upside down can. Say one 5 second blast every 15 or 20 seconds or so. If the machine powers up a lot longer than the minute you are saying it is now then it could well be a temp problem. My bet is on the power supply though. Buy one for 60 dollars or more so you won't be replacing it every 2 years. Also - you may be getting an underpowered supply . The ink to the supply that you purchased looks like it was a good one but things still happen stuff still breaks .
posted by Poet_Lariat at 7:52 PM on August 11, 2011

The thermometer on the CPU is not particularly reliable. One cheap test to see whether your computer is being insufficiently cooled is to just pull the side panel off, point a box fan at the inside of the machine on high, and see if it'll run like that. I personally would not use the coolant from a can of compressed air, as it is essentially liquid nitrogen and I'd worry about it causing some tiny solder joint to crack from thermal stress. Also it has a tendency to cause condensation to form on things, and that's not something you want.

If so, then excess heat is the problem and a cleaning is in order. Try to be gentle with the compressed air, you can potentially mess stuff up that way by driving dust into moving parts, sending small bits of metal (which may have somehow accumulated in the case) flying about, building up static, or various other things. I usually use a rag or my fingers for this job. Make sure to keep yourself grounded with a grounding strap or by holding onto the (plugged in) case of the computer. Compressed air is valuable for any bits you can't get at, but I find there rarely are any.

The above is free and therefore definitely worth a shot, but my bet is also on the PSU (the power supply). If you get a replacement try to either find a model with a decent review or else just look for one that seems reasonably hefty (all else being equal, a heavier power supply will have beefier components) and which lists its maximum capacity based on individual voltage rails (or at least subsets of the rails) rather than just a single number. I agree that about $50 is a good price for the power supply in your average computer. One with a wider fan (120mm or up) will be quieter and have better cooling than one with smaller fans on it, for what that's worth.

If you replace the power supply and the problem isn't improved, no biggie. You now have a spare power supply, which is not a bad thing to have on hand as they do tend to go every once in a great while. Next time you have to check for a dead PSU, you'll have a brand new known-good unit to test with. Unfortunately your next-most-likely culprit is the motherboard. Visible damage is not a requirement here, there are plenty of tiny bits which can get zapped and render you with a wonky board. I'd replace that next, but know that you're rapidly descending into the realm where it could be most anything.

If replacing the motherboard didn't do the trick (and if I were inclined to continue the quest) the next thing I'd try would be replacing the CPU, followed by the boot drive, followed by the RAM (you might try pulling out all but one stick of RAM and trying the computer with each individual stick in turn, to see if maybe you just have a bad stick), followed by the video card I guess. If I had a spare of any of these lying around then I'd try that before I bought anything new to test out.

Welcome to the world of shotgun debugging. Try not to break anything new in the process. Remember to stay grounded, and that eventually the chances of success may cease to be worth the price of continuing. Good luck.
posted by Scientist at 8:22 PM on August 11, 2011

Scientist: I personally would not use the coolant from a can of compressed air, as it is essentially liquid nitrogen

#1. Compressed air FWIW is not liquid nitrogen nor it it remotely at the temperature of liquid nitrogen. Compressed air cans usually are filled with HFC-134a and the exit temp of such a compressed coolant in a can does not remotely approach that of liquid nitrogen. A five second hit on a warm CPU is perfectly save and will neither damage the cpu nor create significant / harmful condensation to a MB.

#2. Using your fingers to clean out a dusty machine instead of compressed air is possibly the worst way on can go about it due to ESD and the possibility of damaging small electronics with big fingers.

#3 Long before you go about replacing a MB you should attempt to re-seat and/or replace RAM . You mention attempting to replace a MB prior to ram. Not correct. The possibilities of a RAM failure are much greater (and much cheaper to diagnose) than a complete MB replacement.
posted by Poet_Lariat at 8:37 PM on August 11, 2011

Yesterday I fixed my ancient iMac's crashing problem by placing a box fan next to it. This model has no place for an internal fan and no side wall of the case to open. One too many thermodynamics lectures convinced me that just a fan fan might waft away the heat nearly as well, which has indeed proven true. Two days of heavy use without a single shutdown, where before they'd happen every few minutes(!) as the day wore on.

Worth a try for now, considering most sources of compressed air are closed at this hour. Good luck!
posted by skbw at 8:46 PM on August 11, 2011

Just got back from the electronics store ... they tested the power supply for free and it was giving perfect outputs (5, 12, 3.3) for over 5 minutes.

I bought a new motherboard but am hesitant to have that installed because it will require a reinstall of Windows, right?

I will try reseating RAM and moving them around in the DIMM slots to see if that makes a difference.

Was kind of hoping it was the power supply because it would at least be an easy fix :/
posted by lilgoyl at 9:21 PM on August 11, 2011

You haven't mentioned your video system. Is it on the mobo or is it a video card? Video cards build up dust on the fan and heat sink fins just like other parts & depending on the model it may be covered with a metal or plastic shroud you'll need to remove to de-dust it properly.
posted by BeerFilter at 9:32 PM on August 11, 2011

It's a video card.

So here's what I've done with the help of my husband ....

Blown out all of the dust
Reapplied thermal paste to CPU
Tested power supply
Tried different video card
Reseated RAM
Move RAM to different slots and tested each individually
Tried a different CPU (we have three computers in our house that are the exact same build)

The computer showed the same symptoms after each one of those steps (shutting off in less than one minute while browsing BIOS)

However, just when I was about to throw in the towel, I reassembled the computer to how it was originally with all original parts and now it's working again. So bittersweet because yay it's working again but who knows why and for how long.
posted by lilgoyl at 11:00 PM on August 11, 2011

If you replace the motherboard, you shouldn't have to re-install Windows. You could try it if the issue re-occurs.*

I recently replaced a motherboard myself, as I was experiencing a similar issue. The PC would turn off without any overheating, and the date kept resetting (i.e. CMOS battery issue). I initially replaced the CMOS battery as well as the memory cards (since they showed countless corruptions in Memtest86), but the new ones then got corrupted as well. I then read online that my particular type of Asus motherboard has a known factory fault which could cause these corruptions. It has been working fine since.

*Note that the CPU socket on the motherboard needs to match the processor.
posted by larb at 4:58 AM on August 12, 2011

>*Note that the CPU socket on the motherboard needs to match the processor.

Actually, any deviation can cause the need for a re-install. Different controller chips, different integrated devices, who knows. If you get the exact same board or one that is very, very simiiar (exact same chipset, same revision) you may luck out. The nice part about XP is that you can do a simple "repair install" and it tends to just work. In Win7 you usually need to install a fresh copy of Windows. I've found Win7s repair install mode thing to not be able to handle new mb's with different chipsets.
posted by damn dirty ape at 9:09 AM on August 12, 2011

This is the motherboard: ASUS M3N78-EM AM2+/AM2 NVIDIA GeForce 8300 HDMI Micro ATX AMD Motherboard

The computer has remained powered on for over 12 hours now. I performed a memtest for ~2 hours and there were 0 errors.

Still would like to figure out what's going on!
posted by lilgoyl at 11:47 AM on August 12, 2011

« Older Help me find the detailed schedule for the...   |   Songwriting variety Newer »
This thread is closed to new comments.