My desktop has been randomly turning off. Not slowly shutting down or bluescreening, but the power will just completely turn off. All fans and lights instantly go off. I then have to unplug it or flip the power switch on my PSU off for a minute or two before I turn it back on. If I try to turn it on without doing that, it lights up and makes noise for about 7 seconds, but then turns off.

What could be the possible causes of this?

PSU failure?
Motherboard failure?
Maybe RAM?

Are there any tests I should run that would test for these?

It's been happening for a couple weeks. Normally it's just if playing a game or watching a video, but today it did it while I was just on facebook.

It sounds most likely to be either a PSU failure or thermal shutdown. Did you recently add any new hardware, or did your system recently sustain any physical trauma or knocking about? The fact that it is more likely to happen when your system is under higher load is indicative that it is either a PSU problem or a thermal problem, but unfortunately from what you said I can't really narrow it down to which one. Can you think of any other details that might help?

No new hardware, but physical trauma is a definite possibility. The first time it happened was right after a 5 hour car trip.

OK, that's good to know. Physical trauma might possibly have dislodged heatsinks, fans, etc. If you have some thermal compound around, it might not be a bad idea to reseat your CPU heatsink/fan, although if that's not too easy, your first step might be running any of hundreds of freeware apps to see at what kind of temperatures your CPU is running.

CPU thermal shutdown is highly unlikely. When CPU's exceed their limit, they downclock themselves, not shutdown the computer. Also, if the CPU is exceeding its thermal limit the motherboard would likely start beeping an alarm and/or there would be system instability.

Of course, something like SpeedFan would easily let you know if it is a thermal problem or not.

Also, RAM, CPU, motherboard, GPU, etc... failures would cause hardlocks and bluescreens, not power loss.

PSU failure is what I would suspect. It is also easy to test. Simply jump the PSU (short the green and black wires on the 24 pin cable) and let it run continuously for 24 hours or so, see what happens. Just make sure you've unplugged the PSU from the motherboard and video card. I'd leave everything else plugged in so that there is some load on the PSU.

However, detailed system specs and temp readings would really help.

Oh, and why are you taking your desktop on 5 hour car trips?

They most certainly do shut themselves off when they reach a certain temp (above that of the slowdown threshold temp). My previous laptop had a GPU kill temp of 110C (which I hit many times due to HP outsourcing their manufacturing to a low-quality company; just google "hp nvidia laptop overheats" to see what I mean.)

Yup, I've run into this many times with Simms. At one point the core machine ("Simms", not to be comfused with "Simms0", "Simm5", etc) had a bad thermal junction between CPU and heatsink, and would randomly shut off and refuse to turn back on until I toggled the PSU switch. Some Arctic Silver took care of the problem.

Here are my speed fan readings:

GPU: 57C
System: 40C
CPU: 82C
AUX: 127C (!?)
HD0: 35C
Core 0: -33C
Core 1: -33C

And my college is 5 hours away. I'm living at home for the summer, but I went to my apartment for a week to pay rent and party. And since I have good internet there (not my dial-up like here...) I brought my desktop so I can hulu and game. (Netbook sucks with streaming video..)

82C is hella high for a CPU; I think my desktop with a Core 2 Quad Extreme idles below 50 and only gets up around 55C under heavy load. Kllrnohj, what do you think?

And btw, 127 sounds like +MAX_INT to me (1 bit sign, 7 bit mantissa) rather than an actual reading.

Oh, and system specs. Not really sure what you need.

Asus P5W DH Deluxe
Core2Duo @ 2.13
Antec PSU 550W

That CPU is definitely running a bit toasty (assuming that's idling); what are the temps on 100% load?

If they exceed slowdown threshold they tend to crash, not kill power. Some motherboards can be configured to shutdown when a certain CPU temperature is reached, but the CPU can't control power itself.

And laptops != desktops. They are extremely different when it comes to thermal management and safety.



Shit man, put that fire out. 82C is EXTREMELY high for a cpu. I have my warnings all set to 65C...

BUT the good news is that it is quite possible that speedfan is simply wrong. This is why knowing the CPU is important. Try this utility and post a screenshot of its output: http://www.techpowerup.com/realtemp/

http://dl.getdropbox.com/u/128496/temp.jpg

Oh, man, this isn't looking good...

EDIT:
I just turned off my computer for ~3min, then turned it back on and checked my BIOS. That said my CPU was 79C, too.

Didn't that say your system fan was off? I think that is a major problem to try to fix.

Seems like someone needs some liquid helium.

No biggy, just buy a new HSF. Depending on the model, they cost between $30 and $80

http://hardocp.com/reviews/cooling/

Find one you like, although you should expect to pay about $50-60 for a solid performer (and they tend to be well worth it).

I would probably go with something like this: http://www.newegg.com/Product/Product.aspx?Item=N82E16835608002

Strong performance, quiet, decent price. It'll also let you push that C2D to well over 3ghz and still keep load temps around 50C

Alright, I'll probably order that one tomorrow morning. What's the deal with thermal compound? Is it worth $10 to pick up a tube? And what do you do with it?

Good god man that thing is HUEG and expensive; I suggest this.



Thermal compound aids in the heat transfer from the CPU to the heatsink (ensures an even surface for the heatsink to bond to.)

You should buy some thermal compund remover to remove whatever thermal compound you already have on your proc, then put on some Tuniq TX-2 (better & cheaper than Arctic Silver 5 IMO) using an old plastic card (credit card, id card, flashcard, etc.), being sure to spread it thinly & evenly and then clamp the heatsink on.

There's an air duct on my case right above the processor, so I should probably get the smaller one. Even then, I still may need to cut some of it out.

When I built my computer, it was missing one of the little screws that go between the motherboard and the case. Since the reviews say that installing it requires a little bit of elbow grease, I'm going to steal a screw from a different computer and replace that while I'm at it. This'll be the first time I've had my motherboard out of my case since I've built it. I'm already getting nervous. I get so paranoid around hardware. :/

Understandably so, but it's better than getting overconfident and then breaking something. I'm glad we were able to properly debug your issues.