Desktop PC powering down shortly after powering on -- what next?
September 2, 2014 8:09 PM   Subscribe

My four year old home-assembled Windows desktop PC was working fine until a recent move. Now, when I turn it on, it seems to be booting as usual but then powers down about 15-45 seconds after powering on. What's the likely problem? What do I do next?

I assembled the desktop in question about 4 years ago (with some help from AskMeFi). Everything was fine until we moved last week. I reconnected everything tonight, and the first time I powered on it seemed fine -- until it powered down about 30s after turning on, at the Windows login screen. I tried a few more times and the behavior repeated -- seemed to be booting as usual, BIOS screen showing normal and Windows loading as normal, but then powering down after 15-45 seconds.

I opened the case and looked for obvious loose cables... all of the fans seem to be working normally...

What's going on? What should I do next? Please hope me MeFi.

posted by Perplexity to Computers & Internet (20 answers total)
Buy a bigger power supply. Yours probably isn't beefy enough. (In the short term, try removing an unessential card like the sound board and see if that changes things.)
posted by Chocolate Pickle at 8:17 PM on September 2, 2014

Does that make sense when it worked perfectly for 3.5 years?

Does inadequate power supply failure mode include working perfectly for 30 seconds and then shutting down?
posted by Perplexity at 8:27 PM on September 2, 2014

Sounds like maybe a thermal event. If the system gets too hot it'll turn itself off, and that can happen right quick.

Check the CPU fan - is it spinning? Is it (and the heat sink) seated properly? Did it get knocked around in the move?
posted by kbanas at 8:32 PM on September 2, 2014 [1 favorite]

I think that if the wall voltage in your new place is lower than in your old place, your supply may not produce quite as much current inside your computer.

And yes, it can take 30 seconds for a shutdown, if some of your drives have delayed spinup.

The way to see is to remove anything unessential, including unplugging every USB device except your mouse (and keyboard?) so as to reduce power draw, and see if you can boot.
posted by Chocolate Pickle at 8:37 PM on September 2, 2014

Does it power down if you interrupt the boot and let it stay on the BIOS screen? Also, make sure that all of your RAM and expansion cards (e.g. video cards, sound cards, etc.) are well-seated, and that your heatsink is still firmly attached to the CPU.
posted by Aleyn at 8:59 PM on September 2, 2014

The move could be coincidence.

One of my office machines had that problem. Once the innards were cleaned of dust, especially around the doovy and whatsit, it worked again fine. Have you cleaned yours lately?
posted by Kerasia at 9:08 PM on September 2, 2014

I would make sure your hard drive is completely backed up before doing anything to the PC with the hard drive still in it. Because your PC won't stay on for more than a minute, you'll have to remove it first, of course, and connect it to another PC (or get someone else to do so).

It is unlikely that the issue you're having will result in damage your hard drive, but I'd do it anyway, just to be on the safe side.
posted by UrineSoakedRube at 9:51 PM on September 2, 2014

Start off by blowing all the dust out of the machine. Canned air is best for this though I've used lung power before. Concentrate on the heat sink blades on the CPU.

Try booting off a CD. If it's good then you've likely got an issue with your windows install. And yes, moving can cause major problems, especially to the hard drive.

Still shutting down? Next step is to disconnect as many things as possible to eliminate a peripheral issue (ram, hard drive, usb devices, keyboard, mouse and video card if you have one). Start it up without anything and see if it continues to power down after a period of time. If it stays on start plugging things back in one at a time (RAM first, video card, then HD). When the issue reappears you've found the cause.

If the computer shuts down even after removing everything then you've probably got a motherboard or power supply issue.
posted by talkingmuffin at 10:14 PM on September 2, 2014

I'm not thinking underpowered power supply, but bad power supply. The fact that it's actually booting up to the login screen kind of, at least IMO, rules out RAM or anything along those lines as the issue.

I'd swap the PSU, then boot into the BIOS and let it run until it shuts down(does it?) watching the CPU temp if your BIOS shows that.

I can't imagine it being a thermal thing unless your heatsink is not actually seated on the CPU anymore. And at least when i've seen that, it would shut down much more quickly the second time... not at the exact same place at the login screen.

I'd next suspect a bad motherboard or video card, and that booting into windows is causing a piece of hardware to either power on, or switch from it's default low power no drivers state(IE "VESA" video on GPUs) to a higher power normal state and then it craps out.

I'll also note that the recent times i've had issues like this after a move... the shitty power in my new place was actually damaging my hardware, and had created this issue in the first place. I went through several power supplies and several motherboards before i figured this out. Good times!
posted by emptythought at 10:17 PM on September 2, 2014 [1 favorite]

Definitely boot into bios and watch the temperature meter before doing anything else. If it's going up and up until it shuts down, you know that it's a thermal issue, which I'm guessing would be the heat sink coming loose somehow. I think the main key with troubleshooting at first is to boot it into bios so you can eliminate some of the other factors which might be causing it. If it stays in bios just fine for a long time, but booting into windows causes the crash, that tells a very different story than if it crashes in bios after 30 seconds.
posted by markblasco at 10:52 PM on September 2, 2014 [1 favorite]

You might try removing and reapplying the thermal compound between the CPU and its heat sink. If that connection gets compromised, the CPU will overheat quickly even when its fan is working.
posted by qxntpqbbbqxl at 11:36 PM on September 2, 2014

My first impulse is definitely swap the power supply, so seconding that.
posted by brennen at 1:18 AM on September 3, 2014

Something got moved and a fan isn't working perhaps. But n'thing power supply. Always buy the best power supply possible - which fortunately is about $40
posted by mattoxic at 5:23 AM on September 3, 2014

Given 4 years since assembly, plus the move, my first suspicion would also be the CPU heat sink / thermal paste / CPU fan as a cause for this problem. Especially if you have a stock CPU fan, the mountings for those can be pretty awful at the best of times. Thermal paste also degrades over time.
posted by protorp at 5:28 AM on September 3, 2014

If I were at a customer's house dealing with a PC with those symptoms, here's what I'd do:

1. Take the side panel off the case.

2. Unplug the power cord, then press and hold the front panel power switch for at least 10 seconds to discharge the power supply and make sure it's no longer supplying sleep-mode power to the motherboard. If the mobo has an LED on it to indicate the presence of sleep-mode power, pay attention to that.

3. Blow the CPU heatsink clean with canned air. Whoosh a bit into the power supply as well to see how much of a dust cloud can be raised. Give the case a general blow-out to get rid of as much dust and loose fluff as possible. If the computer is in a smoker's house, canned air might not be enough to de-gunk the CPU heatsink, in which case I'd reach for the stiff-bristled artist's brush I keep for that purpose (it fits between the fan blades, so I don't have to waste time taking the heatsink assembly apart).

4. Unplug the SATA power and data connectors from all hard disk and optical drives.

5. Remove all back-panel connectors and plug-in cards, including graphics cards and RAM sticks. The only things I want left connected to the mobo are the CPU, power cables and internal front-panel cables. If there's anything plugged into external connectors on the front panel (USB sticks etc) remove those too.

6. Clear the CMOS settings using the jumper or battery-removal method.

7. Refit the power cord and attempt to start the PC.

At this point I would expect to hear the typical continuous BEEEEEEEEEEEEEEEEP that a PC makes to complain about a complete lack of RAM. If I don't hear that, the fault is narrowed down to CPU, mobo, or (most likely) power supply. But I almost certainly would hear that.

8. Repeat step 2.

9. Carefully inspect the edge connector on one of the RAM sticks. If the gold plating shows even the slightest evidence of discoloration - especially if that's in a pattern that could plausibly be a system assembler's thumb print from four years ago - carefully polish the gold plated fingers with a white pencil eraser. Don't use an ink eraser, they're too abrasive.

You can rub quite hard with the eraser, and you should get some satisfyingly black streaks appearing on the white eraser surface that your efforts are busy chewing away. You do need to be precise, and very careful not to dislodge any of the tiny chip-style components close to the connector fingers.

Afterwards, use canned air or a well-practiced non-spitting mouth to blow off all rubber crumbs, then re-inspect. You should find that the connector edge you've been working on is markedly brighter than those you haven't.

And of course I didn't need to tell you that at no stage in this process must your own fingers be allowed to come into contact with the edge connector fingers, even though it seems I just have.

10. Use canned air to blow out all the RAM slots, and all the expansion connector slots, and the fan/heatsink assembly on any now-disconnected graphics card.

11. Refit the cleaned stick in the RAM slot closest to the CPU. Be absolutely sure that it's pressed fully home into its slot: if it is, the retainer clips will snap fully into place of their own accord without you needing to nudge them.

12. Plug the keyboard into the back panel. If the computer has inbuilt video, connect the monitor to that. If not, clean the graphics card's edge connector the same way as the RAM stick's, then refit the graphics card and connect the monitor to that.

13. Repeat step 7.

At this point I would expect to hear a normal power-on self-test success beep, and for the computer to start up and then complain about invalid CMOS settings. If it still does the long no-RAM BEEEEEEEEEEEEP, or fails to do that even though it managed to after step 7, the RAM stick is most likely bad.

14. Repeat step 2.

15. If there is more than one RAM stick, repeat step 9 with the next one. If step 13 says the RAM stick you've already plugged in is bad, replace it with the one you've just cleaned; otherwise fit the next stick into the next RAM slot. Pay attention to slot colors and whatever rules your mobo has about the order you need to fit sticks to make dual-channel or triple-channel modes work.

16. Repeat step 7.

17. All being well, get into the BIOS setup screen, start by loading optimized defaults, then tweak as necessary. At the very least you'll need to reset the date and time. I also like to get into the power settings, change the front panel button settings from "delay 4 seconds" to "instant off", and set Restore After AC Loss to "always off". Sometimes you also have to turn something like Smart Fan Control on so that the PC doesn't sound like a jet taking off.

18. Restart the PC to make your BIOS settings stick.

19. It's probably fixed by now. To gain confidence, leave it sitting in the BIOS setup menu for a good 15 minutes while you chat with the owner and drink their coffee, then check the PC Health settings to make sure the temperatures and voltages all look reasonable.

20. Repeat step 2.

21. Reconnect all the SATA cables. Power it back up. Make sure it boots up normally and behaves itself; there's a 95%+ chance that it will do. Edge connectors are easy to manufacture, but they're still the least reliable part of a desktop PC, especially since so many technicians don't seem to be able to control an urge to fondle them before assembly.
posted by flabdablet at 6:22 AM on September 3, 2014 [1 favorite]

Super happy fun time note about the instructions above(which are solid, by the way);

Windows vista and newer will refuse to boot with a cleared bios and either just instantly reboot or give a "windows failed to start please wait while the troubleshooter runs" screen if clearing the bios switched a SATA hard drive/SSD out of either AHCI or IDE emulation(which it never should have been on in the first place!). If it fails to boot after clearing the bios, flip that setting!

First damn thing I check if the bios has been cleared(or I clear it, or I suspect the settings have been munched) and the system won't boot for any reason now.
posted by emptythought at 10:46 AM on September 3, 2014

Thanks for all of the fantastic answers, especially flabdablet.

After getting a chance to work on it a little more tonight, I have update/questions.

The interior was very dusty including all of the fans, and the CPU fan was loose/disconnected. The heatsink was truly filthy. So I cleaned everything up with canned air and remounted the CPU fan, and now I'm running just fine. I feel reasonably confident that the issue was just CPU temperature, although I could obviously wrong.

But! I didn't reapply any thermal paste or anything like that, and I suspect I should maybe do so and/or replace the fan. I downloaded a temperature monitoring thing and it's running around 50C at very low load. Internet results are mixed but it sounds like this is bad, right? It's a Core i7-870 if that matters.

Should I just get "thermal paste"? What is that exactly? Should I replace the fan? Here's a photo of the CPU fan/heatsink which I assume is totally standard.

And a screenshot of the RealTemp screen measuring the CPU temp. How bad is this?
posted by Perplexity at 8:29 PM on September 3, 2014

If the heatsink wasn't getting proper airflow due to a fan that was either disconnected or mounted improperly, then the problem was almost certainly that the CPU was getting too hot. Most CPUs and motherboards will attempt to throttle the CPU to cool it, and failing that it will shut down completely to prevent damage to it.

I have an i7-920, and I just replaced the stock heatsink and fan with a new one (the fan was dying on the old one). HWiNFO64 tells me that the CPU is running at between 35-50C right now, but before the fan died on my old one, I'd regularly see it at 50-60C with the stock cooler. You should be fine. If you regularly see temperatures above 80C, or the thermal throttling is kicking in (like it was on mine before I replaced the heatsink), then you should look into a new heatsink and fan. If you're really worried, you might try running Intel's processor diagnostic and see what it tells you.

TL;DR: you're fine.
posted by Aleyn at 10:08 PM on September 3, 2014

it's running around 50C at very low load. Internet results are mixed but it sounds like this is bad, right? It's a Core i7-870 if that matters.

Nah, the first gen core-i CPUs ran SO hot compared to the core2duos, or anything that came after them like sandy bridge. I had an i7 920 desktop like Aleyn, and my work still has a bunch of i7 860 machines. They both idle hot, and run pretty damn hot under load. I eventually got a ginormous highly regarded/reviewed heatsink and nice thermal compound... and it didn't make much of a difference in idle temps.

I would probably file this under the category of "fuck it" unless it keeps shutting down.

I'm amazed that just dust was enough to make it shut down that quickly though, wow. Is it the stock heatsink/fan combo?
posted by emptythought at 3:21 AM on September 4, 2014 [1 favorite]

Some mobos have a BIOS setting you can use to make them shut down if they don't see the fans moving. There's typically a delay on it so you can't instantly shut down a PC just by briefly jamming a fan, but if fan speed shutdown is enabled for a fan that's not actually connected, it will typically kick in well before an unventilated heatsink has had time to get dangerously hot.

Thermal paste goes between the bottom of the heatsink block and the top of the CPU. The stock cooler that comes with a boxed CPU will typically have slightly too much thermal paste pre-applied, so if you assembled your own PC without knowing what thermal paste is, that's probably what yours is using. Over four years it will have dried out some. The main effect of that is to make it behave more like a rather weak glue than like the grease it started out as. The heat's carried mostly by the paste's filler particles anyway, not by its oily carrier, so drying out doesn't affect thermal performance much.

If the PC got bashed in transit hard enough to knock the fan off the heatsink fins, there is some possibility that the rather heavy heatsink block has also been briefly knocked away from the top of the CPU, resulting in a crack in the old dried thermal paste (it's held down by spring-loaded latches that do let it move some, which it has to in order to avoid cracking the CPU through thermal cycling).

But as long as the fin block hasn't managed to shift sideways much, the most likely outcome of that even with quite nasty old thermal paste is that it will all ooze back into place over time.

Replacing the paste is the right thing to do if you're taking the heatsink off the CPU entirely. If you've no need to do that because it never actually overheats (and 50°C is pretty typical for a grunty CPU with a stock cooler) then don't bother touching it.
posted by flabdablet at 9:54 AM on September 4, 2014

« Older Wading through   |   Guardians of the Galaxy Release schedule Newer »
This thread is closed to new comments.