[SOLVED] Hard shutdown with no warning

[talsemgeest]

There is quite a serious problem with my computer. Quite randomly, it will suddenly shut off, almost like a power cut but with power still supplied. Most often when I try to turn it back on, all that happens is the hard drive light flashes once and the fans in the case twitch, before turning back off. After switching off the power for a few minutes it will usually allow me to turn it back on.

I am very sure that it is a hardware error, as it happens both on windows and ubuntu, and after a reinstall of both os's. I have also reset the bios settings to the defaults, changing only what is necessary, so I am reasonably sure the problem does not lie there, however I am not 100% sure.

After the hard shutdown, I have felt the different components to rule out overheating. Case temperature is only a few degrees hotter than the external temperature, CPU feels well below 40, HDDs feel a bit hotter, possibly 45 degrees C.

The shutdown seems to be reminiscent of when my graphics card failed, being an instant shutdown with the computer refusing to start until the graphics card was removed.

What I really want to know is if there is any way of narrowing the problem down to a specific component, and find out what the specific problem is and how to fix it. All help/comments are welcome and appreciated.

Just in case it helps, here are my system specs:

AMD athlon 4200+
Asus m2a-vm hdmi motherboard
2 x 1GB DDR2 667 RAM
Creative sound blaster audigy OEM
Asus P7131 Hybrid tv tuner card
Asus DVD RAM drive
Floppy drive
40GB IDE HDD with ubuntu 8.04 installed
160GB WDC SATA HDD and 320GB Seagate SATA HDD set up with fake raid) to make:
320GB RAID0 with vista installed
160GB in RAID0 with itself as a backup HDD

Thank you in advance for your help.

[coffeecat]

After switching off the power for a few minutes it will usually allow me to turn it back on.

It could be many things, but this sounds like the CPU overheating.

CPU feels well below 40

How long after the shutdown did you feel it? If you mean that the heatsink on the CPU cooler didn't feel particuarly warm, that doesn't necessarily mean the CPU isn't too hot. If you mean that you can feel the CPU directly, do you actually have a CPU cooler? What type is it? Is the fan working?

Go to Synaptic and install one or both of 'computertemp' and 'gkrellm'. The first is a Gnome panel app, the second a stand-alone app and both will give you a readout of CPU temperature if your motherboard supports this. See what temperature you are getting just before shutdown. Also, check in the BIOS to see what temperature is selected for autoshutdown - I know you reset to defaults but it would be useful to know what the actual tripping temperature is.

[jonabyte]

I'd have to agree with the above, probably the cpu overheating as it will definitely cause the system to shutdown without warning.

It could also be the power supply, I have seen that in the past as well.

[talsemgeest]

How long after the shutdown did you feel it? If you mean that the heatsink on the CPU cooler didn't feel particuarly warm, that doesn't necessarily mean the CPU isn't too hot. If you mean that you can feel the CPU directly, do you actually have a CPU cooler? What type is it? Is the fan working?

Sorry I wasn't more specific, I just felt the heatsink, and that felt only slightly warm. I am using the stock cpu cooler, with some aftermarket thermal paste.

The fan that is on the heatsink is controlled by the motherboard, with the speed lowered when the temperature is low, and is sped up when the cpu gets hotter. The speed hasn't notably sped up so I assume the temperature being reported by the cpu is well within limits.

As for it being the power supply, is there any way of confirming that, other than buying a new psu?

[coffeecat]

As for it being the power supply, is there any way of confirming that, other than buying a new psu?

Have you got a PSU in another machine you could try out? Or could you borrow one? I can understand you not wanting to go to the expense of buying one and finding it makes no difference.

Back to the CPU temperature though. With the fan behaviour you describe I agree this is unlikely to be the problem, but try one or the other of the apps I suggested.

Another thing is to look in /var/log/messages. Do this after an unexpected shutdown and check the messages for the minute or so before it shut down. There might just be a clue there.

How about faulty memory? Except that if the machine was behaving OK until recently this is unlikely.

[talsemgeest]

Unfortunatly the only other psu I have has too low of a power rating, so that is out of the question, and I don't think my friends would want to go without a psu for a day. But even then, the problem seems random, I can't reproduce it when I want to and so I can't say for sure that the problem is still there or if it is fixed.

As for the programs you suggested, computertemp stayed at 40 degrees, no matter what the load which means that it is not properly detecting the temperature. As for the other one, should I start it from the command line, because I don't think it is on the menu.

I think I will run memtest tonight to make sure it isn't a memory related problem.

[coffeecat]

As for the programs you suggested, computertemp stayed at 40 degrees, no matter what the load which means that it is not properly detecting the temperature. As for the other one, should I start it from the command line, because I don't think it is on the menu.

That's a pity about computertemp. That's possibly an incompatibility with the data coming from the motherboard, because it works fine for me in a couple of machines, but in an older one it tells me the temperature is fluctuating rapidly between -20C and 140C - or something like that.

Sorry - I forgot that you don't get a menu entry for gkrellm. I had to add one myself. It's simply 'gkrellm' from the command line. Right-click on the upper right corner of the gkrellm window, or away from the data areas, to do general configuration. CPU temp is not switched on by default - you'll find that under options > builtins. And you'll probably need to increase the width of the gkrellm window in the main options pane.

Having said that, I believe gkrellm relies on the same kernel sensors as computertemp, so you may get the same result. I only suggested both apps, because gkrellm is more comprehensive than computertemp, but it uses more screen area.

[talsemgeest]

I am going to be switching back to windows for a bit to play some games, so I will use the utilities that came with my motherboard to monitor the temperature from there.

[talsemgeest]

I have noticed that the temperature of my cpu is higher than it used to be. It used to average around 35, going up to 40 under load. Now it averages 45 going up to 60 under load. It seems that it is more likely to happen when under load (i.e. encoding video, playing games, watching 1080 h.264 streams...) but still havent been able to get an accurate temperature reading just before shutdown.

Also, once it has shutdown it is getting more and more unlikely that pushing the power button will turn my computer back on. I have ruled out my pci devices and my hard drives being the problem by unplugging them and then seeing if my computer will turn on, but none of them did anything. It certainly seems like a heat problem as the longer I leave my computer off the more often it will turn back on. I am also starting to see it turn on, get to the os loading screen only for it to shut down there.

This is effectively driving me crazy. All help is very very welcome.

[talsemgeest]

This morning it shut down at a temperature of around 36, so I dont think that the problem is entirely cpu overheating. Unfortunately that last shutdown has killed my computer. I have taken out all the parts except my motherboard, PSU and cpu, and the exact same thing happens when I try to turn it on (Fans twitch then nothing). When I take the cpu and turn it on the same thing happens again.

So I am guessing it is either that my cpu is dead and because of that my motherboard doesnt pick it up, or my mother board is dead.

Any ideas/opinions?