View Single Post
  #2  
Old December 11th 04, 12:45 PM
Paul
external usenet poster
 
Posts: n/a
Default

In article , wrote:

On Fri, 10 Dec 2004 05:26:44 -0500,
(Paul) wrote:

In article ,
wrote:

Asus P4P800 Deluxe -



---snip

Wow, thanks for all the tech info paul, I can definately use much of
that information for further annalysis.

I have monitored the 1.5V load thru the bios and with Asus probe, both
seem to indicate similar readings. I cannot be assurred which one is
more accurate, however.
But...both indicate a drop of only .12V without load and of course
this fluctuates to minor positive incriments as well. I wish I would
have tried the voltage adjustment in the bios when the working
replacement card #1 was running so bad in this board... I have seen
this setting. I probably will try it anyways...

Now, just to add the monkey wrench, the northbridge does seem warm. In
fact, during my trouble shooting, I could feel the warmth, not
neccessarily heat, comming from my open chassis. This was pinpointed
to the northbridge chip. Incidently, I ordered 2 stock P4 intel boards
for back up measures and the northbridge heatsinks are twice the size
of the Asus heatsink.


But in relation to:

The way
that works, is when operating in 8X mode, the AGP lines are
terminated with load resistors. The resistors cause the
signal voltage to drop to 0.8V, while the I/O pads driving
them are still actually powered by 0.8V. In a sense, this
is still powering by 1.5V, but you'll still see references
to 0.8V, even though the circuit is not fed by 0.8V.


This point may be the clue, but...

Is this feature done on mobo or on the card ?
If it is done on mobo, then if the load resistors are in fact not
dropping the voltage to .8V, what would that mean ? Unless I am miss
interpreting, this would definatly shorten the life of any AGP 8X card
in this board running at 8X...

I must also say that there is no AGP setting in this bios which allows
me to set the AGP "_X" feature, so it is automatic and sets at the
optimum...I guess.

Thanks so much for the feedback.
I am attempting to RMA the board and the new warranty video card is
going in the intel board. I just like to take these hardware issues as
far as I can, technically speaking...

Regards,

ñíñjà¤têç


I didn't think AGP voltage registered on the monitor chip.
Would that be Vcore you are looking at ?

The terminators could be inside the GPU. I cannot really be
sure, as the AGP standard does not address how the AGP
I/O is actually built by companies like ATI and Nvidia.

The termination resistors are there to make a clean looking
trapezoidal signal. The fact that the signal swing is reduced,
does a couple of things. It can reduce the emissions caused
by the switching noise, and a reduced swing signal switches
faster. The termination scheme, in fact, speeds up the signals
two ways. The amplitude is reduced, so the signals don't have
to go as far. And, a parallel termination type, also removes
the RC delay associated with series damping type terminations.
But that is a tiny number, like 250pS or so. (And with the 8X
rate, the clocking methodology is probably such, that delay is
immaterial anyway. I haven't read up on how the clocking works.)

To try to detect the heat, you would need to run the motherboard
first in 4X mode, measure the Northbridge temperature, then
run it in 8X mode. Otherwise, it will be tough to tell the
difference between the two standards.

If I was looking at this, I would need to get out a
multimeter, and find a place to monitor the AGP voltage
on the motherboard. I would get a copy of the AGP standard
and find out which pins on the socket have the AGP I/O voltage
on them. Then, when the motherboard is powered up, probe the
holes along the edge of the motherboard connector, for the pin(s)
in question. But, you have to be very careful when doing
stuff like that, as one slip could short the conversion circuit
for the AGP, and burn it out. (It may not be protected, in the
interest of saving money for Asus.) You should only try this
with the motherboard sitting on the test bench. Clip the multimeter
ground onto a connector lug on the I/O connectors, so the
ground probe doesn't get in the way.

As the AGP standard is no longer on the Intel download site,
I use this one. Page 50 has the pinout, and the AGP voltage
is on the pins labelled Vddq1.5

http://web.archive.org/web/200303140...0_final_10.pdf

Page 63 shows a picture of the signal amplitude (0.8V)
compared to the voltage used to power the I/O pad (vddq1.5).

On page 64, you can see a FET being used on the receiving
chip, as a resistor to ground. That FET, is how the chip
is able to digitally control the terminator. The terminator
is not always enabled, and would be disabled when the card
is set to 4X mode. In that case, the signal swings to 1.5V,
and the power for the I/O pad is still 1.5V. (If the AGP
I/O is bidirectional, there are probably terminators in
both the Northbridge and the GPU - two instances of the
circuit, pointing in different directions.)

As far as the Northbridge temperature, other things that
affect it, are running the DDR RAM at DDR500. Or boosting
the DDR voltage, as the voltage used to run the memory,
also runs the Northbridge. The Northbridge and memory
DIMMs share common I/O signals, and generally that means
the I/O power for the two devices will be common as well.

Even dual channel versus single channel, makes a marked
difference in the operating temperature.

Looking at the AGP I/O signals themselves, would be
pretty hard without some fancy test equipment. The AGP
I/O runs at 533Mb/sec in 8X mode, and that means a
5GHz digital sampling scope would be needed, for single
shot capture. You would probably end up capturing an
eye diagram for the I/O, rather than trying to track
the 1's and 0's. Ah, the good ole days...

HTH,
Paul