Rapid Unplanned Overheating

[waterlubber]

CUST-BUILT PC was destroyed due to overheating!

Okay, it's not destroyed, but my computer just overheated on me, or the PSU failed.

What should I do? I waited 15 or so minutes before rebooting.

At the time, I was recording KSP and was transitioning from the craft to the KSC.

Advice?

Warnings?

Your PSU is ready to explode?

[Camacha]

CUST-BUILT PC was destroyed due to overheating!

What exactly did it do? Your description is not too clear

[Brethern]

My money is on either one or more fans are clogged with dust or not working, or your power source isn't good enough to keep up with the requirements of ksp and recording at the same time.

[Mazon Del]

You should download one of the various free temperature monitoring programs that are out there. They can log what temps things operate at, and depending on your gear and the software, it might be able to tell you which items are operating close to their red lines.

[Camacha]

Generally a PC is protected from overheating. It just shuts down instantly when a certain temperature is reached. This temperature can often be set by the user. If that happens you should check for dust, airflow and whether all fans are in proper working order. Setting the temperature higher could solve the problem, but make sure not to exceed the maximum temperature for your specific chip. These can be found on the manufacturer website.

We still don't know if that's what happened, though.

[LaytheAerospace]

Unplug the PC and take out the CMOS battery on the motherboard for a few minutes and try to turn it on.

The clear CMOS jumper would be faster. At any rate, that's not going to help anything unless this is the result of him screwing with the settings on his motherboard.

Anyway, if your PSU was failed, you probably wouldn't be able to boot up at all. I've seen PSU failures where it would only go down once load started to increase and the PSU wasn't able to maintain steady power, but that kind of thing is pretty rare. If you can get into Windows, chances of it being your PSU are pretty low. Overheating seems most likely right now.

Go grab SpeedFan or HWMonitor and watch your CPU temps. Depending on what CPU you have, you'll hit the shutdown point somewhere in the vicinity of 90C-105C, though you can start running into weird issues before that, and running the chip near its thermal limit for a long time will cause degradation (not likely to be noticed unless you're overclocking, it eats up the headroom). Under load, 40-60C would be considered good, but it depends very much on what chip you have. Intel chips run cooler than AMD chips, and newer chips usually run cooler than older chips.

You can also cut KSP out of the loop and just run a CPU stress test to see if you overheat and/or crash. Prime95 is the go-to tool for that, just run the default settings for at least 15 minutes and watch the temperatures. They should level off pretty quickly. Prime95 will tell you if your CPU starts producing garbage.

[Camacha]

Anyway, if your PSU was failed, you probably wouldn't be able to boot up at all. I've seen PSU failures where it would only go down once load started to increase and the PSU wasn't able to maintain steady power, but that kind of thing is pretty rare. If you can get into Windows, chances of it being your PSU are pretty low. Overheating seems most likely right now.

Very strange things can happen when the PSU is dying, including the described behaviour, but I would indeed first look at temperatures and later at the PSU.

[waterlubber]

First, I run Linux so most software may not work. My fans are clean as I had recently built it, but what may have happened was that the case was opened and the PSU cooler was drawing air the wrong way. I doubt the CPU because when it shut down, my speakers and USB devices went, because the mobo keeps the USB on 24-7 (bit of a pain as I need to unplug the speakers every night because they have accent lighting). However, the temperature in Hardjnfo I think is the mobo and not CPU. The CPU is not overclocked, it is an AMD FX4300.

I think I might remove the PSU and attempt to place loads on it, so if it does fail it doesn't unleash the 9001 Volt Kraken.

[Camacha]

Okay, I feel the descriptions are all a bit vague, so it's time to set things straight:

- What happened exactly? You say it overheated, but how do you know? Describe what you saw, heard, smelled, et cetera.

- You talk about your speakers and USB devices that 'went'. What do you mean by that?

I think I might remove the PSU and attempt to place loads on it, so if it does fail it doesn't unleash the 9001 Volt Kraken.

What, you are just going to poke your PSU until it gives up? If you are going to tinker with it, your best bet is replacing it with another to test whether the same problem occurs.

[waterlubber]

It judt randomly shut off. The fans were running rather fast at the time. My speakers and USB things shut off because they lost power. I was recording and the computer stopped. I thought I accidently hit the Reset button, so I tried rebootjng, but the system wouldn't reboot until I waited a bit.

[steve_v]

First, I run Linux so most software may not work.

lm-sensors.

Also, ACPI values in '/sys/class/hwmon/hwmon*/device/'

[LaytheAerospace]

First, I run Linux so most software may not work.

Use lm_sensors: https://wiki.archlinux.org/index.php/lm_sensors

My fans are clean as I had recently built it

#1 problem when people have issues like this after building a computer is not seating the cooler correctly on the CPU. If it's not seated evenly, or if you used too little/too much thermal paste, then you can very easily overheat under load while being able to handle light tasks without any obvious problems.

Before you do anything else, you need to check your CPU temperatures. Once you verify that it's an overheating issue, try reseating your cooler. Wipe away the old thermal paste when you do, the stuff isn't really reusable after it's been baked on.

but what may have happened was that the case was opened and the PSU cooler was drawing air the wrong way.

Your PSU produces very little heat, and isn't usually responsible for the main exhaust. You could pump the PSU exhaust through the front panel of a healthy computer with no effect. Similarly, the side panel might impart a single digit temperature swing. Unless your computer was already very badly overheating, these small changes can't cause your computer to shut down.

I doubt the CPU because when it shut down, my speakers and USB devices went, because the mobo keeps the USB on 24-7 (bit of a pain as I need to unplug the speakers every night because they have accent lighting).

Any number of things could cause the power to drop out on USB after a shutdown. This isn't really evidence of any particular failure mode, unfortunately.

I think I might remove the PSU and attempt to place loads on it, so if it does fail it doesn't unleash the 9001 Volt Kraken.

No offense, but if you're coming here with these kinds of questions, you're not going to be able to test a PSU beyond a power on test. Please go check your temperatures before you damage your hardware trying to test something you don't understand.

It judt randomly shut off. The fans were running rather fast at the time. My speakers and USB things shut off because they lost power. I was recording and the computer stopped. I thought I accidently hit the Reset button, so I tried rebootjng, but the system wouldn't reboot until I waited a bit.

Sounds like classic overheating. Never seen a PSU need a cooldown time after failure to boot back up. The computer just doesn't draw very much power for a few seconds, so even a PSU limping along on its last leg can usually get enough juice to start trying to POST. A CPU that's reached its thermal limit, however, isn't going to do a damn thing until it's cooled down enough.

Also, it would be helpful if you linked us to your PSU's model. There's an enormous difference in quality from brand to brand and unit to unit. There are tons of $20-$30 PSUs out there looking to trick a novice builder into wasting their money on something that will blow up in a month. I've got a pile of them in the corner.

Edited November 10, 2014 by LaytheAerospace

[Camacha]

It judt randomly shut off. The fans were running rather fast at the time. My speakers and USB things shut off because they lost power. I was recording and the computer stopped. I thought I accidently hit the Reset button, so I tried rebootjng, but the system wouldn't reboot until I waited a bit.

Really sounds a lot like the overheat protection. It is not very typical for PSU failure, so I would check out the temperatures, like has been said a couple of times now.

[Nuke]

if there are any whining or hissing sounds comming from the psu, then you might have some dieing caps in there. if it were me id replace them. but this is one of those high voltage jobs that requires someone who knows what they are doing. normally people just get a new one.

No offense, but if you're coming here with these kinds of questions, you're not going to be able to test a PSU beyond a power on test. Please go check your temperatures before you damage your hardware trying to test something you don't understand.

you also just cant use a multimeter here. you need to test things under load since voltage will change as you draw more current from the rails. voltages may be fine when there is nothing drawing current, as soon as you try to pull an amp or two, voltage drops out of spec. there are dummy load boards out there for this kind of thing. but again not for electronics noobs.

Edited November 11, 2014 by Nuke

[Camacha]

if there are any whining or hissing sounds comming from the psu, then you might have some dieing caps in there. if it were me id replace them. but this is one of those high voltage jobs that requires someone who knows what they are doing. normally people just get a new one.

He could have a look inside with a bright light. If any caps look swollen or burst that's trouble all right. It seems unlikely the PC would resume normal operation after cooling down though, caps generate other problems.

[Maxwell Comp]

Use lm_sensors: https://wiki.archlinux.org/index.php/lm_sensors

#1 problem when people have issues like this after building a computer is not seating the cooler correctly on the CPU. If it's not seated evenly, or if you used too little/too much thermal paste, then you can very easily overheat under load while being able to handle light tasks without any obvious problems.

I came here to say exactly this. If it's overheating, a probable cause is a lack of thermal paste. I have some Artic Silver that I used when re-seating my old computer's CPU heatsink and it's pretty good, but really almost any thermal paste will do the job if you apply it correctly. It's hard to get a bad paste, and the high-end ones perform only a few degrees better than average ones when they are both applied correctly. A kit may come with a cleaning solution to remove the old paste, and you should use a coffee filter or microfiber cloth to wipe off the CPU and heatsink contact points. Don't let any dust or fingerprints get on the contacts, because that will decrease it's performance.

[MartGonzo]

It judt randomly shut off. The fans were running rather fast at the time. My speakers and USB things shut off because they lost power. I was recording and the computer stopped. I thought I accidently hit the Reset button, so I tried rebootjng, but the system wouldn't reboot until I waited a bit.

99% sure that's heat related, I had a similar issue, where the CPU would get too hot and shut itself down whenever I fired up a game.

The problem turned out to be a lack of thermal grease between the CPU and cooler, so it only cost a fiver to fix.

If the computer boots up again after resting for a few minutes it's almost certainly heat related, hopefully it will be something simple for you too, check there is good coverage of thermal grease under the cooler and that the cooler is seated right. (I've had that problem too, always double check when re-attaching the fan that it's flush with the motherboard.

Oops sorry, just read the other replies and realised I'm just repeating what others have said.

[LaytheAerospace]

you also just cant use a multimeter here. you need to test things under load since voltage will change as you draw more current from the rails. voltages may be fine when there is nothing drawing current, as soon as you try to pull an amp or two, voltage drops out of spec. there are dummy load boards out there for this kind of thing. but again not for electronics noobs.

Yep. What's important is the voltage under load. Does it droop? Ripple? Ring? Unless you happen to have a PSU test harness, or an oscilloscope and the components to build your own dummy load, then testing isn't really possible. You can get a PSU tester for ~$20 that will let you plug in the 24 pin and a molex connector, but this kind of testing is useless for a multi-rail design. Not only will you not be able to draw the PSUs total output, but if loading up one rail causes another to droop, you'll never see it. I also have a hard time believing one of those cheap devices would handle a high end PSU that can dump 60A+ onto a single rail.

With all this in mind, here's my step by step PSU testing process.

1. Short green pin to ground. Did it turn on? If not, throw it in the garbage.

2. Swap suspect PSU into another computer and stress test. Does it crash? If so, throw it in the garbage.

3. It's not the PSU. Put it back in the original computer and move on.

[Camacha]

It's worth mentioning that having a proper A-brand PSU is always worth the trouble. Cheap units, even when operating normally, tend to cause all sort of trouble and vague problems and can damage your expensive hardware. Subpar PSU's often are not even capable of delivering their supposed wattage. Rather than providing their power over a sustained period, they can only push out their rated power for milliseconds, spelling trouble as soon as you actually load it. A-brand power supplies will happily deliver everything promised and often a lot more for extended periods of time.

I consider the PSU the most important part of a computert, together with storage (and backups, which is part of storage). Those two things have to be done right. It is no use buying a fancy processor when you haven't gotten the basics right.

[LaytheAerospace]

It's worth mentioning that having a proper A-brand PSU is always worth the trouble. Cheap units, even when operating normally, tend to cause all sort of trouble and vague problems and can damage your expensive hardware. Subpar PSU's often are not even capable of delivering their supposed wattage.

At the risk of derailing this thread further with talk of PSUs, here's a good way to evaluate a PSU if you don't recognize the brand, or want to compare within the same brand.

Check the PSU's specifications for how many amps on the 12V rails, multiply by 12 and that's how many watts the PSU can provide where it matters, because the power hungry components all pull from the 12V supply. The closer this number is to the PSU's rated wattage, the better the PSU is. Cheap PSUs will inflate their power by making tons of amps available on the 5V and 3.3V rails, which it won't be utilized.

For example, I'm using the SeaSonic X650 Gold, a 650W power supply. It has 54A on a single 12V rail, for 648W, or 99.7% of what was claimed. Cheap PSUs will often deliver 60-70% of the rated power, and if you try to draw more, then it's not going to handle it very long, if at all. If the PSU manufacturer is going to such great lengths to deceive you, it's not because their PSU is a fine piece of equipment that will last you years and years.

[Camacha]

If the PSU manufacturer is going to such great lengths to deceive you, it's not because their PSU is a fine piece of equipment that will last you years and years.

Exactly, this was what I was trying to convey. The same can probably be said for inflated SSD performance numbers, but let's not get into that discussion here