System loses power, how to diagnose?

[TopherC]

A homebuilt computer of mine has some kind of fault that causes it to shutdown suddenly. I'm not sure how to diagnose this problem.

A few months ago I had this computer working fine. Then at one point it began having problems coming out of sleep mode. Most of the time, after being in sleep mode, it would not wake (with a keyboard press), and the power button would do nothing. I could boot it up by switching off the power in the rear, waiting a couple minutes, switching it back on and then hitting the power button. I soon learned not to use sleep mode (running Windows 7) but to either leave it on or simply shut it down.

Still, this bothered me so I got a newer power supply. This fixed the problem entirely, so I was happy for several weeks. Just now it has begun acting strangely again. Sleep mode is again a problem as before, recoverable by powering off in back for a few minutes. But now after turning it on, the computer will frequently shut itself down by suddenly losing power, often within a few minutes of turning on.

I'm an experienced system builder, but I'm not sure what to try at this point. The only thing I can think of is that there may be a short that develops with a certain amount of heat. Is it a mounting/standoff problem? Since this seems to be related to sleep mode, I'm more suspicious of the motherboard than I am of other power-consuming peripherals. But it's not practical to replace the MB, CPU, and RAM just to test this theory until I have more confidence in it.

Any suggestions?

Many thanks,
Topher

 

[banthracis]

Have you run Memtest on the RAM for at least 8 passes?

Is there an OC on the build?

Have you Prim95 12 hrs min to test CPU stability?

Have you made sure no malware/virus on system?

Are drivers on everything and OS updates all up to date?

DL HW monitor and get some load and idle temps.

What sleep setting do you have enabled in BIOS?

 

[TopherC]

I will check on these when I get home. I have run memtest86, but not more than one pass since this problem started. I'm not overclocking anything. I haven't run Prime95 yet on this system, but will do that next along with temp monitors. I've been running this with the case cover off for the past few weeks. I'll have to check the sleep settings in BIOS.

What seems most significant to me is that I have to switch the power off completely at the power supply for a minute or two before I can even get a flicker of life out of it again after one of these spontaneous shutdowns. Has anyone else seen this kind of behavior?

 

[JackNaylorPE]

I'll echo what banthracis said tho 1st thing I'd check is CPU temps.

What size / model PSU is in use ?

Anything in Event Viewer ?

 

[AndyC682]

Im experincing something similar.

The only difference is, mine restarts itself after a few seconds.

My cpu, gpu, mobo temps rarely get any higher than 50c

 

[TopherC]

I'm sorry that I haven't updated this thread for a few days. After running lots of tests the system became incapable of passing POST last Sunday. I've ordered a new MB, CPU, and RAM now on the assumption that the motherboard was bad. I'm still curious if others have any suggestions for possible points of failure. But at this point I can't run any more tests (unless I find my POST decoder card). I'm reporting these investigations in case they might help others.

I have an Antec 550 W power supply in there right now, and had a 380W PS earlier. The 550 is overkill. I don't have the model of MB -- It's either an MSI or Gigabyte board. The CPU is an Athlon 64 3700+.

I ran an overnight memory test with memtest86, which completed 8 passes (over 3GB RAM) and found only one bit error on the 2nd pass (7th test). I wasn't happy with that, but don't think memory corruption is the main issue.

I also ran SpeedFan and watched temperatures while running Prime95 for two hours (under Windows). The system was stable, and CPU temperatures never exceeded 45 C. Other temp sensors showed reasonably cool temperatures throughout the test also. It seems like if the system can stay up for more than about two minutes, it will remain up for several hours at least and withstand stress tests, benchmarking, and gaming. I stopped Prime95 after 2 hours because I was looking for a more immediate kind of problem.

Since this problem began, I've only had a Windows session working three times. It has powered off a few dozen times while trying to boot. A few times I could actually log in only to have it power down seconds later. But usually it would power off within 10 seconds of beginning to load Windows.

I tried swapping out the video card with an older spare (both NVidia chipsets), and saw the same problem. My video card driver was out of date, but since trying to update this I haven't been able to get the system up and running fully. Booting into safe mode has the same problems.

I had noticed earlier that when booting up off of various live Linux CDs I never experienced a spontaneous power failure. So I installed Ubuntu 10.04 on a spare drive, wondering if loading off the HDD rather than CD would make a difference. I was able to boot into this OS several times, without once hitting a power failure! Alternating between this and Windows, I was still not able to boot fully into Windows. The hardware doesn't like Windows?? However, this installed the Grub bootloader and I noticed that as soon as the Grub menu appears during a boot, a kind of screaching sound would come from somewhere inside the computer and it would power off after about 6 seconds. But as soon as I hit a key that made a difference (up/down/enter, but not shift or ctrl) this sound would stop and the system would be stable at the Grub menu. I tried this in the dark, and saw no sparks. It seems that there are some particular set of operations that trigger this problem, probably a short circuit somewhere. Ubuntu Linux does not trigger this, but Windows and Grub do at various times.

I tried several BIOS settings and couldn't find any adjustments that fixed the problem. Finally last Sunday after running more tests, the computer refused to pass POST and has been non-functional since.

 

[jsc]

Why is everybody neglecting to consider the PSU?