[ILUG] [OT] computer fails to post

[Declan Moriarty]

Recently, Somebody Somewhere wrote these words
> Quoting Declan Moriarty <junk_mail at iol.ie>:
> 
> > Recently, Somebody Somewhere wrote these words
> > >
> > > So I'm left with the problem of determining if its the CPU
> > > or motherboard.
> >
> > Can you get into the bios? Do you see the ramtest? Does the
> > screen power on? Have you tried a minimal boot - reset button
> > unplugged, 1 small Dimm, 1 Drive, no pci, no usb etc.
> 
> no video signal from the machine, even though the graphics card
> is working. So basic assumption is that something is failing
> during POST. All pci cards were disconnected.

A lot of the post is after that powerup. The only items before post
are to see if the thing can power up.

It also fingers the psu.
There's a 'Power Good' line which should go logic low when the
board is to be switched on. I gather you have a voltmeter. You can
check that for a logic low level (well a switch of levels) on (I
think) the grey wire? Measure against the black. If you find it
difficult to make contact, stick pins or needles through the
wires. Sites give you the lowdown, e.g.

http://xtronics.com/reference/atx_pinout.htm

There is also some link you can make to force a powerup. Your
hardware, your choice. It's on the PSU, IIRC.

If you use pins, MAKE SURE THEY TOUCH NOTHING! Remember the 
Socerer's Apprentice :-).

> 
> > If you can verify it failing POST (i.e. starting POST) it's
> > the motherboard - with 90% certainty. If the cpu is dead it
> > does nothing. Like if it powers the monitor on, I'll bet on
> > the motherboard.
> 
> monitor remains on standby even when power is on in the graphics
> card and the various fans start and the single hdd spins up.

It sounds like you need a cpu, if you have a 'Power good' behaviour. 
You need to know, however, that Power Supplies give out their rated
current, and if more is asked of them, they collapse the voltage,
affecting Power good. Check the Value on each line, particularly
the orange 3.3V I would substitute in a power supply, not try yours
in some old box. A PSU giving 10Amps on one line can go faulty in
that it shows overcurrent after 5 amps. It's been known.

> 
> 
> > Another idea is to 'receive' a signal from it. The amd-64 runs
> > at gigahertz. That's divided down all over the place AGP, PCI,
> > IDE, etc.  Now you can feed in a signal from an aerial into
> > whatever receiver you have, as all these things broadcast. A
> > steady whine is what as well as the cackle. You can also hear
> > a harmonic of it.  Get an electrically quiet area, and hold an
> > aerial or rabbit's ears about a foot away from the box.
> >
> > You'll get the power supply with a transistor radio on long
> > wave.  IDE/pci will be at the low channel numbers on a telly
> > (The old Band 1 for 405 line transmissions). The higher
> > numbers (over channel 21) are uhf, i.e. channels 21-68 at
> > 470-900Mhz You should pick up a harmonic of something there
> > Just wavy lines or a buzz on the sound.

The cpu  (on a telly) now becomes highly important here. I bet you
won't get it.

> >
> > If the above fails....
> >
> > If you can get a multimeter, set the voltmeter for 5V, and
> > measure (CAREFULLY) between pins 28 & 14 on the flash (if it's
> > 28 pin) or pins 16 & 32 if it's 32 pin for 5V. 14 is the
> > ground, and keep the black lead there for further
> > measurements. Take care not to slip from pin to pin - that's
> > the way to blow things.
> >
> >
> > It's numbered this way
> >
> > 	__    ___ Pin 1|  \_/  | Last pin Pin 2|       | |       |
> >
> >
> > Pin 14 is ground. The falsh, or eprom with bios in it, is
> > usually big enough to measure on. Pin 22 (on a 28 pin) or pin
> > 24 (on a 32 pin) as Address Latch Enable, and that should be
> > clocked regularly while the post is running. It will read
> > between 2 and 3 volts if clocked.  Likewise the pins on both
> > sides of 14 (or 17) are the databus, and they should vary. If
> > they don't it inclines me to condemn the cpu (60/40)
> 
> Didn't think of using a voltmeter, cheers!

I was thinking afterwards, the voltage depends on the pulse width,
so don't be fussy about 2-3 volts.
> 
> > By this stage you'll be so fed up you won't care and will
> > gladly pay for a new m/b and cpu :-D.
> 
> and if I do it will be a motherboard with the seven segment
> led's so I know which component just went toasty. :P
> 
> 
> Frustration just isn't the word :(
> 
> 
> "Nothing's foolproof to a sufficently talented fool"

-- 

	With best Regards,


	Declan Moriarty.