Around the time of the P4B motherboard (S478/SDRAM), a circuit was
added to the AGP slot, called "AGP_Warn". (AGP_Warn is the word
printed right next to the circuit.) The logic for AGP_Warn says:
IF (AGP_Card_Inserted & TYPEDET_Sez_3.3V_only_Video_Card)
THEN stop_computer_from_powering_up
That circuit prevents some really old, incorrectly keyed video
cards from blowing up a 1.5V only Northbridge with 3.3 volts.
The first round of implementations, included a Red LED next to the
AGP slot. If the Red LED lit up, you then knew that a bad video card
was being used. (And you also knew for sure, why the computer would
not power up.)
In the second round, the LED was removed to save cost. The word
"AGP_Warn" could still be seen printed next to the AGP slot, so
you knew the circuit was still there. You could not absolutely
and for sure, know it was there, because the LED was not populated.
There were still holes in the board for the LED.
I had a look at my P4C800-E, and I really don't see any signs
of the circuit. It could still be there, but just not placed on
the motherboard in the same way as was done previously. (When the
transistors for the circuit are spread around the board, it won't
be possible to recognize the form of the circuit block any more.)
In terms of the implementation, pg.50 of this doc shows the pinout.
TYPEDET# signal is on pin 2A (I believe that is on the back side of
the card). A proper modern video card should ground pin 2A. With
an ohmmeter, you should see a direct short between pin 2A and 5A.
http://web.archive.org/web/20030314013601/http://developer.intel.com/te
chnology/agp/downloads/agp30_final_10.pdf
Now, some manufacturers have what are called "testability engineers".
At the factory, these guys make sure the circuit board designs are
modified, so that the circuit board can be tested. It appears that
some of them have added a resistor to the video card, that runs from
pin 2A to ground. If, when using the ohmmeter, you read a reasonably
small but finite resistance (say 50 ohms to 150 ohms or so), it
could be that was added for testing. But, what such a resistor can
do, is foul up the Asus "AGP_Warn" circuit. Also, in cases other
than Asus, I understand the inclusion of that resistor, has caused
some motherboard AGP regulators, to produce voltages inappropriate
for running the logic - AGP is supposed to be 1.5V or 3.3V, and a
video card with a resistor on pin 2A, can cause the Vreg on the
motherboard to deliver 2.x volts instead, which is good for
neither purpose.
So, before giving anyone heck, I would grab another AGP card and
try it in the AGP slot. Not a 3.3V only card of course (it won't
fit in the slot anyway).
If another AGP card works, but the PNY continues to stop it, you
can return the PNY for your money back (as an RMA is only going
to give the same problem). If you own a multimeter, you could
take a stab at measuring the resistance between pin 2A and 5A
on the video card, to see if there is a dead short (as defined by
the standard), or a low value resistor (as defined mistakenly by
test engineering).
Just a guess,
Paul