Desktop randomly shuts off

[MrSweet1991]

Hi All,

So I visited a customers home and I was told it would randomly shut off, instantly. So after a little look I took one of the two sticks of RAM out and it started up, to verify the problem I shut off the computer and stuck the second RAM stick in, it wouldn't turn on. So for me it would appear that one of the sticks of RAM was faulty. However because the computer wouldnt boot properly into windows I took it home to have a proper look, and when I took the computer back to the customer I said, "because of how slow the computer is with 1GB DDR, it may be worth just seeing if it will boot with the second stick again" and to my surprise it did, before leaving I said if you have anymore problems just pop this second stick of DDR back out and it should be fine.

Two weeks later he said he's having the same problems, he said he took the one stick of RAM out but even then it would still randomly shut off. From that stage one of two things would happen, the CPU fan would spin for a fraction of a second and shut off or the computer would turn on.

I've taken the computer back home and this time I'm going to attempt to test the memory with memtest, and look at the PSU and Motherboard.. But that's the problem, I could get another PSU and leave that in to see if it will stay on, but what about the motherboard, or maybe even a dying CPU which I've heard could cause the issue.

Is there a better approach of diagnosing this issue? And assuming the PSU seems to be okay and memtest sees no issues, how can you test a MOBO or CPU other than replace them with known working ones?

 

[Joe Khamis]

In the past I have booted off a copy of Stresslinux(1). Hiren's Boot CD also has a nice selection of DOS and Windows based stress utilities. (1) http://www.stresslinux.org/sl/

 

[MrSweet1991]

In the past I have booted off a copy of Stresslinux(1). Hiren's Boot CD also has a nice selection of DOS and Windows based stress utilities. (1) http://www.stresslinux.org/sl/

Hi Joe,

Thanks for the swift reply, in respects to the stress tests, can this help in determining if the CPU is dying and not anything heat related? And what tools can be used for MOBO testing?

Thanks.

 

[Joe Khamis]

I use Memtest86+ (check RAM) and Prime95 (check CPU errors). There's also a Video Memory Stress Test. To be honest, I cannot recall having to try the latter. As far as a test for the motherboard specifically, sorry, no suggestions there.

 

[Digital Micro]

I would test the memory just to be sure it also doesn't have problems, but I would consider it the power supply. You can get a basic power supply pretty cheap. If it's anything more (motherboard or cpu) it's not worth repairing. I am assuming this by you saying it has DDR memory not DDR3 or 4. If it turns out it continues to shut off I would just give them a credit instead of a refund and move their data over to a new PC at no charge.

 

[FremontPC]

Also, if this is DDR, we're talking a fairly old system. Check the tops of the capacitors to make sure none of them are bulging or oozing, that can give you some pretty weird results. If it's AMD, make sure it's got compatible memory.

 

[MrSweet1991]

Hello all,

Thank you very much for your replies, I do have an old PSU to hand but I was curious if there's a more methodological approach to diagnosing this sort of issue. I did think to myself if may be heat related, but my only doubt was if it was getting towards the thermal limit, wouldn't the OS start to go really sluggish just before it hits the thermal limit and shut off?

And yes, it is DDR1 and is an old system, checking the capacitors is something that I also think is worth checking so that's a good point.

• Right I'll do some stress tests and monitor GPU and CPU temperatures to look for anything suspicious.
• Failing that I'm going to disassemble everything and just have a quick visual check to make sure no one (computer shop) have installed extra standoff screws which could short the board etc,
• then I'll run memtest to test the memory and hopefully it will get through the test wihout shutting off
• From that point I'll use a working PSU

I'll report my findings, thanks to everyone who replied its appreciated

 

[ComputerRepairTech]

Which processor is it? Chances of it being cpu are quite slim. You mentioned something about you heard it could be a dying cpu but in reality it could technically be a lot of different things but vast majority of the time its psu or motherboard related. I can't say i've ever seen a random shut off being caused by ram though. You mentioned GPU, what kind of gpu is in the machine?

Is the shut offs really random? Like can they occur as soon as the windows login screen comes up? Perhaps they only happen while hes browsing media rich sites? etc etc

Edit: No tech replaces motherboard before PSU unless they have a reason to believe its motherboard like maybe they know for sure the PSU is good or they see a bulging capacitor on the motherboard. You generally do not want to put in new hardware in a system with a potentially bad psu.

 

[Moltuae]

Memory and motherboard faults usually cause BSODs and reboots not sudden shutdowns. I say usually because there are exceptions but, unless there's an overheating issue, it's unlikely the fault lies in the motherboard or memory modules.

If it turns off very intermittently/randomly, I would say almost certainly a PSU fault. In fact, I'd go further and say most probably a bad/dry joint in the PSU. A very common fault in my experience.

Change the PSU; I bet it fixes it.

 

[brandonkick]

I wouldn't waste a lot of time on a DDR system, that system is probably at least 10 years old.

After 10 years, who knows what has failed or is about to fail. I'd put in a *new* power supply
and give it a spin for a week or two. If that doesn't fix it, then they need a new computer. I don't
know what your rates are, but by the time you try to hunt down what all failed / is failing and fix
it, they could have bought another much newer system and pay you to migrate their data.

But I guess to answer the question I think you were answering....

Often times the computer will give you clues..... beep codes or certain things happening a certain way
and so on. First thing I always do is make sure the inside of the system is clean, and that the fans are
all spinning. If the system won't post, or isn't stable then I remove all but one memory stick. I'll try to boot
and repeat the process trying all available memory slots until I get a booting system that appears to be stable
with just that one memory stick. If I go through all the slots with that one stick and I can't get it booting or stable
then I'll go to one of the other sticks if there are any, or I'll swap in one I have lying around. It's not guaranteed to be
compatible, but it's worth the shot as long as the voltage and densities are the same. If I get no love there, as a last
ditch effort I will try to swap in a new power supply and strip the system down to the bare essential hardware needed
to boot to the BIOS. If I can get that far, I'll add hardware piece by piece until something stops the system from posting
or causing it to be unstable.

That much trouble is usually only worth it on new systems that are on the upper end of the price scale. You can sink 4-6 hours
into a system like this easy and might not have a concrete answer as to whats wrong. For the cost of the time spent alone,
your getting into the area of where the customer could buy a brand new middle of the road system. Worst case scenarios can
have them spending more to fix the machine than what it's worth.

 

[cypress]

Pop in a new PSU like everyone else is stating. Make the customer aware that you are going to need it on your bench for a couple of days to see how it runs. Put it through the loops with stress software and see what occurs. My bet is on the PSU as well.

 

[MrSweet1991]

Hello Technibbles =]

I just want to say a huge thank you to each and every one of you who have put the effort in to give me guidance and past experiences to help me fix this computer system. I will at some stage reply to you all individually as well just so you know I do read all of them.

So in the little time I've spent with the system I've come to some findings, first off the 7600GS graphics card reached 98 Degrees Celcius running Unigine Valley for approximately half an hour with the side panel off, with the side panel on it *may* have achieved the 125 Degrees thermal shut off as the majority of the time he just watches tons of Netflix.

Secondly, I'm quite sure two capacitors next to the CPU are bulging at the top, I'm going to strip it down so I can see it at eye level to verify it is actually bulging, this could very well explain the "random" shut offs.

Using HWInfo I've monitored the 3, 5 and 12 volt rails under stress and (assuming the readings are right) they all regulate at around the correct voltage)

 

[FremontPC]

Good work. Check for those next to the PCI slots too, especially any that might be hidden by the graphics card. Also eyeball the caps on the graphics card itself, just in case.

 

[MrSweet1991]

Hi all,

So I've stripped the PC and here are some photos of the offending capacitors.

And just thought I'd show this professionally applied thermal paste using the "Wall of China" method

 

[brandonkick]

Two more quick things:

Capacitors do not necessarily need to swell and crown at the top to be bad.
More than just the ones that are visibly leaking / crowing very well may be bad. Only
way to know for sure is to desolder one leg and test their ESR. Again on a system this
old i don't know that I'd bother but just know that replacing the swelled ones doesn't
mean the system is in tip top shape cap wise.

Secondly, even though that was a horrendous job at putting on the thermal paste.... chances
are it's long dried out since it was applied. Most often you'll find the OEM used a thermal pad
and they do not usually ever need to be replaced, at least not before the system should be.
I guess the long and short of this one is, even if they did it right.... it would still probably need
to be redone. But replacing it and doing it the right way can help. Too much thermal paste starts
to act like an insulator of heat and not a conductor, and that will shut down / kill a CPU real fast.

Closing comments on the GPU: See if it has heatsinks that you can replace the thermal compound on
or maybe try googling to see if anyone has created heat sinks for it. I've seen it done with various things.
Also try looking at the case ventilation setup. Maybe thing about replacing any 80mm fans with 120mm
fans where you can, they are not only quieter but push much more air. If the system only has one fan
see if you can add at least one more.

 

[MrSweet1991]

Two more quick things:

Capacitors do not necessarily need to swell and crown at the top to be bad.
More than just the ones that are visibly leaking / crowing very well may be bad. Only
way to know for sure is to desolder one leg and test their ESR. Again on a system this
old i don't know that I'd bother but just know that replacing the swelled ones doesn't
mean the system is in tip top shape cap wise.

Secondly, even though that was a horrendous job at putting on the thermal paste.... chances
are it's long dried out since it was applied. Most often you'll find the OEM used a thermal pad
and they do not usually ever need to be replaced, at least not before the system should be.
I guess the long and short of this one is, even if they did it right.... it would still probably need
to be redone. But replacing it and doing it the right way can help. Too much thermal paste starts
to act like an insulator of heat and not a conductor, and that will shut down / kill a CPU real fast.

Closing comments on the GPU: See if it has heatsinks that you can replace the thermal compound on
or maybe try googling to see if anyone has created heat sinks for it. I've seen it done with various things.
Also try looking at the case ventilation setup. Maybe thing about replacing any 80mm fans with 120mm
fans where you can, they are not only quieter but push much more air. If the system only has one fan
see if you can add at least one more.

Thanks very much for your reply, in respects to the capacitors I have read quite a few contradicting points about them. One thing for certain is, a capacitor doesn't have to bulge to be broke and some have said a bulging capacitor (as you've stated) doesn't mean its broke. But then this whole techy talk got going then where some was saying you cant always test a capacitor because it may not run in parallel or something? Regardless they're bulging and one has definitely leaked so I'd say it needs to go

As for the graphics card its actually a passive cooling one, so it relies on the case circulation to cool it down. What I was surprised at (given it was a brand new pre built system) is that the only other fan is a little 80mm fan at the front.. The CPU fan is isolated with a fan surround to go straight out the side. But I have some IC Diamond and some remover and surface prep (from the Arctic Silver kit) so I'm expecting a good drop in temps. I mean, he's had it from new and for many many years, so it did run absolutely fine with the cooling it has now, with the exception of the thermal paste drying out so I think it will be interesting to see what a re paste could achieve for that GPU.

All in all, thanks for all your input and I may look into how to properly test capacitors as one day I may come across something a little more substantial than a 775 board where its worth replacing the capacitors.

 

[FremontPC]

and some have said a bulging capacitor (as you've stated) doesn't mean its broke.

I wouldn't subscribe to that point of view. I think Brandonkick was saying that even if you replace an obviously bad one, that doesn't mean the board is fixed. The guidance I've read states that if a cap is bad, all of the caps of the same make/value/voltage should be replaced with new, quality caps. Some even offered "kits" for popular mobos back when this was a raging problem.

As you mention, this system isn't worth the trouble anyway. Hosed for sure.

 

[Moltuae]

But then this whole techy talk got going then where some was saying you cant always test a capacitor because it may not run in parallel or something?

I suspect what was meant by that was when a capacitor is in-circuit. Most components can't be reliably checked in-circuit. Electrically neighbouring components can have an influence on readings and component values, giving false positives or negatives. Without being able to see what the component is connected to or studying a circuit diagram, it's difficult to be absolutely certain that the readings you're getting are correct. Most of the time you can make educated guesses and assumptions though. For example, if the measured value of a resistor is very close to it's marked value, you can probably safely assume it's ok, especially if there are no visible signs of damage. To be 100% certain though, you need to measure components out-of-circuit and that is especially true for capacitors.

 

[MrSweet1991]

Thank you both for the clarification, so to avoid the whole issue with false readings due to neighbour parts and interference from other things on the same circuit, you can just remove the capacitor and test it on its own?

Also, a bit off topic but does anyone have some feedback on PSU testers? I do have a multimeter somewhere that was given to me as well and I was wondering if that would do a better job?

 

[mr m]

A bulging or swollen capacitor is an imminent sign of failure. Don't trust it even if it appears to be working. I have a client whose mobo has had swollen caps for about a year. I explained to him what's coming his way and told him to keep good backups since he didn't want to replace the mobo.

Moltuae is correct about testing a cap out of the circuit. Caps store electricity so be careful. Testing caps are tricky by their nature. You must short them out because they are designed to hold a charge. The charge will dissipate over time in an unenergized circuit.

The larger the cap the more energy it stores and the bigger the bite. Small mobo caps won't hurt you but a large cap like the motor starting caps in you garage door opener can knock you on your butt so be careful.

Google how to test a cap if you're interested. It's much harder than testing a simple resistor.