Hi! I have an vexing problem with powering up - sometimes it takes twenty or thirty tries to get my box to 'stay' on.

Here are the details. First the box: AMD Semperon 3400+ 2G RAM, new 500W p/s with 2 huge fans, APIC loaded, new BIOS battery installed as well. The O/S: Linux Mandriva 2008, single user.

Now the problem: The unit will power-up with correct POST beeps, begin to load the O/S, then, at random stages, will suddenly completely shutdown with no warning. Other times, it will load correctly, and I'll be 'puting away (it's running sometimes 3 hrs, 1/2 hr, or 2 mins) and suddenly it just dies. Now one other thing that is really odd: after twenty or thirty tries, every so often, the power-on and reset buttons become functionless. I have to turn of the UPS before they will work again. I've examined the dmesg file and other kernel/system log files and nothing untoward is reported there (aside from the sudden power down message).

I've tried disabling APM, APIC, and a host of other tidbits that have to do with power control or shutdown. No dice. I get the feeling it's a dud m/b.

Any ideas? I'm stumped here. Rather not have to buy another m/b . . .

Could be a HW problem. One thing that can have those symptoms is a bad cxn between cables or m/b and daughter boards.
Make sure its all off, then take the cover off and re-seat all the boards and disconnect/reconnect all the cable cxns. Gives a nice fresh full cxn, and eliminates that possibility.
Do this to external cxns as well ie mouse/keybd/video.

Also eliminate the UPS by trying it without the UPS.

This also sounds like it could be a power supply problem. I've experience the same issues with a bad power supply. Eventually, it just failed entirely.

I agree with chrism01 and Mr. C. Definitely try reconnecting everything again, insuring they all are good. Then I'd get a new Power Supply. Sometimes these things will do some crazy things with absolutely no signs that anything is wrong. I had one spark and smoke once. Destroyed my mobo. Sorry. Don't mean to worry you.

Ah well, if you let the magic smoke out, you've had it
http://dictionary.die.net/magic%20smoke

Im having a similar problem where my computer reboots continually first thing in the morning (when its cold), after about 5 trys it eventually stays on.

I took out the graphics card and although it still needs a couple of restarts first thing, the computer seems to be more stable. I think its power supply in my case.

Classic cold-fault scenario. When a computer starts up, a small crack in a track on eg m/b cxn to cpu may cause issues, once it warms up the metal expands ever so slightly (closing the gap) and you're good to go.

ah interesting

thanks Chris

I'm inclined to think it's the m/b with 'cold fault' syndrome. Chalk up another triumph for 'el cheapo' technology - wave bath soldering! This is my third m/b, fourth power-supply, etc. . . .

[ Secretly, I'm hoping the newly reported DNS flaw will eventually shutdown the Internet so I can dispose of my computers completely - it would be the perfect excuse to get out of admin work and retire ]

Btw, my computer fired up first time this a.m. -- and it's cold in here -- go figure!

Thanks for the ideas!

spud_linux

Well, if you did what I said in post #2, that could happen ... for a while ... then you're back to square 1 ...

These kind of faults are difficult to pin down exactly, unless you either have spare HW to swap in/out and/or you want to start hooking up an O-scope or a Logic Analyzer.

(yeah, I did HW fixes for a job for a while...)

After the last multi-try startup, the dmesg file reported:

"APIC error on CPU0: 40(40)" 2406 times!!

After a successful startup, the error count for this message drops to 1044. Interestingly enough, I can consistently get it to startup first try by holding the 'soft touch' power-up button for three seconds (the manual mentions it must contact for 3 ms.)

The plot thickens.

One thing for sure -- I'm making backups, and saving work-in-progress very frequently!