Incompatibility between Intel P4 MB and ATI Radeon Graphics...

Was Intel MB designed not to support ATI Radeon model graphic cards?
I tried 3 Radeon cards on Intel D845PT P4 and none of them work perfectly.. The MB manual said "D845PT board is only compatible with 1.5 V AGP cards. My card l think is all 1.5V 8x/4x/2x but I don't know for sure. All 3 Radeon graphic work with AMD MB.

Anyone knows more detail on the incompatible issue between Intel P4 MB and ATI Radeon card?

wrote:
Was Intel MB designed not to support ATI Radeon model graphic cards?
I tried 3 Radeon cards on Intel D845PT P4 and none of them work perfectly.
The MB manual said "D845PT board is only compatible with 1.5 V AGP cards.
My card l think is all 1.5V 8x/4x/2x but I don't know for sure.
All 3 Radeon graphic work with AMD MB.

Anyone knows more detail on the incompatible issue between Intel P4 MB and ATI Radeon card?


There are a couple things to consider.

An obvious one, would be if the AGP digital interface on
a motherboard, was non-compliant, or wasn't capable
of working at high transfer rates.

There is no evidence Intel chips ever suffered from this.

I have one motherboard, with an ALI AGP chipset, where
that seemed to be the problem. That's how I became a
fixture here, when I spent three weeks trying to get it
to work, and nobody could help me. That was one of the
few motherboards, which was a dismal failure for me.
And it seemed to be a chip design issue on the motherboard
(ALI Northbridge). The chip simply wasn't fast enough on
the I/O.

*******

A second kind of failure, is related to TYPEDET. That's
a simple logic signal, that sets the voltage on the
AGP bus interface. TYPEDET is supposed to be grounded
for 1.5V cards. A motherboard slot circuit, has to switch
between 1.5V and 3.3V when commanded to do so. And TYPEDET
is the signal controlling that power selection.

There were a few card designs, where some idiot designer
used a 50 ohm resistor to ground, instead of a good solid
ground. (Someone discovered this on a Matrox card, and
some factory engineer probably told the design engineer
to do it for "design for test" reasons.) The motherboard
designer on the other hand, assumed when doing their transistor
level switch, to work with a solid ground as an indication.
On boards which used 50 ohms to ground on TYPEDET, the
motherboard AGP slot voltage might be around 2V or a bit more.
This isn't what the chips on either end were expecting, and
could lead to problems. If, say, the switching threshold
doesn't have an automatic threshold adjustment.

(Slightly more details from my answer back in 2007...)

http://www.techiehq.net/pc-hardware/...5g5-63235.html

To check that 1.5V powered video card is properly designed,
you check for a zero ohm reading between pin 2A and 5A.
Rather than say, 50 ohms.

There are two supply voltages, but three signaling levels.

http://www.playtool.com/pages/agpcompat/agp.html

On that page, you can see the cards keyed for 1.5V or 3.3V.
But the signaling levels are 0.8V, 1.5V, and 3.3V. A way
to make a smaller amplitude signal, is to do something like
a double termination. And that can allow an interface with
a 1.5V power supply, to end up with 0.8V amplitude signals on
it. As far as I know, no "true" AGP 8X cards were made, using
0.8V pad drivers and 0.8V signals. The cards continued to be
powered by 1.5V power source. I don't know any of the details,
of how they made the chips work on that interface (switchable
termination or whatever).

As far as I know, my 845 board didn't cause me problems.
It might have had an old ATI 8500 or the like in it. I don't
know if I ever tested my 9800Pro in there. I have tested
my 9800Pro in my 440BX motherboard (for fun) - it should
have worked, based on slot keying, but the screen would
not light up and I had to remove it. The card continued
to work just fine after that, no damage.

The only card back then, where any concerns were raised,
was ATI 9700, and I don't remember the details of what
was said at the time. (It could be, there was no way to
qualify the card when it was made, so every motherboard
it was plugged into, would be "a new test case".) That's
the only card I might be tempted to do more Googling for.
Everything else should have worked.

And I don't think the practice of using a 50 ohm resistor
for TYPEDET is that common. It might have been seen on
a couple Matrox cards. Early in the life of AGP,
there were a few failures related to gimpy 3.3V supplies
on the AGP slot, and that was a motherboard issue. After
about TNT2 or so, that stopped being a common issue observed
by users. I doubt your slot is "winking out" on an overload.
That just doesn't happen any more.

The only other really bad mistake, was the SIS305 cards keyed
for both voltages, when the card was fixed at 3.3V. Those
could damage a 1.5V AGP slot, and it is because of that card,
that "AGP Protect" circuits were added to 845 motherboards
like the one you're working with. My P4B 845 motherboard
revision 1.05, was the first revision of that motherboard,
to be protected against SIS305 cards or similar (improperly
keyed cards). A red LED on the motherboard, checks TYPEDET,
and if no ground is present, that indicates a damaging
SIS305 could have been plugged in. The protection circuit,
prevents the ATX power supply from being turned on. And
the red LED glows as a hint as to how you screwed up.

Paul

On Tuesday, September 17, 2013 7:48:13 AM UTC-7, Paul wrote:
wrote:

Was Intel MB designed not to support ATI Radeon model graphic cards?

I tried 3 Radeon cards on Intel D845PT P4 and none of them work perfectly.

The MB manual said "D845PT board is only compatible with 1.5 V AGP cards.

My card l think is all 1.5V 8x/4x/2x but I don't know for sure.

All 3 Radeon graphic work with AMD MB.



Anyone knows more detail on the incompatible issue between Intel P4 MB and ATI Radeon card?





There are a couple things to consider.



An obvious one, would be if the AGP digital interface on

a motherboard, was non-compliant, or wasn't capable

of working at high transfer rates.



There is no evidence Intel chips ever suffered from this.



I have one motherboard, with an ALI AGP chipset, where

that seemed to be the problem. That's how I became a

fixture here, when I spent three weeks trying to get it

to work, and nobody could help me. That was one of the

few motherboards, which was a dismal failure for me.

And it seemed to be a chip design issue on the motherboard

(ALI Northbridge). The chip simply wasn't fast enough on

the I/O.



*******



A second kind of failure, is related to TYPEDET. That's

a simple logic signal, that sets the voltage on the

AGP bus interface. TYPEDET is supposed to be grounded

for 1.5V cards. A motherboard slot circuit, has to switch

between 1.5V and 3.3V when commanded to do so. And TYPEDET

is the signal controlling that power selection.



There were a few card designs, where some idiot designer

used a 50 ohm resistor to ground, instead of a good solid

ground. (Someone discovered this on a Matrox card, and

some factory engineer probably told the design engineer

to do it for "design for test" reasons.) The motherboard

designer on the other hand, assumed when doing their transistor

level switch, to work with a solid ground as an indication.

On boards which used 50 ohms to ground on TYPEDET, the

motherboard AGP slot voltage might be around 2V or a bit more.

This isn't what the chips on either end were expecting, and

could lead to problems. If, say, the switching threshold

doesn't have an automatic threshold adjustment.



(Slightly more details from my answer back in 2007...)



http://www.techiehq.net/pc-hardware/...5g5-63235.html



To check that 1.5V powered video card is properly designed,

you check for a zero ohm reading between pin 2A and 5A.

Rather than say, 50 ohms.



There are two supply voltages, but three signaling levels.



http://www.playtool.com/pages/agpcompat/agp.html



On that page, you can see the cards keyed for 1.5V or 3.3V.

But the signaling levels are 0.8V, 1.5V, and 3.3V. A way

to make a smaller amplitude signal, is to do something like

a double termination. And that can allow an interface with

a 1.5V power supply, to end up with 0.8V amplitude signals on

it. As far as I know, no "true" AGP 8X cards were made, using

0.8V pad drivers and 0.8V signals. The cards continued to be

powered by 1.5V power source. I don't know any of the details,

of how they made the chips work on that interface (switchable

termination or whatever).



As far as I know, my 845 board didn't cause me problems.

It might have had an old ATI 8500 or the like in it. I don't

know if I ever tested my 9800Pro in there. I have tested

my 9800Pro in my 440BX motherboard (for fun) - it should

have worked, based on slot keying, but the screen would

not light up and I had to remove it. The card continued

to work just fine after that, no damage.



The only card back then, where any concerns were raised,

was ATI 9700, and I don't remember the details of what

was said at the time. (It could be, there was no way to

qualify the card when it was made, so every motherboard

it was plugged into, would be "a new test case".) That's

the only card I might be tempted to do more Googling for.

Everything else should have worked.



And I don't think the practice of using a 50 ohm resistor

for TYPEDET is that common. It might have been seen on

a couple Matrox cards. Early in the life of AGP,

there were a few failures related to gimpy 3.3V supplies

on the AGP slot, and that was a motherboard issue. After

about TNT2 or so, that stopped being a common issue observed

by users. I doubt your slot is "winking out" on an overload.

That just doesn't happen any more.



The only other really bad mistake, was the SIS305 cards keyed

for both voltages, when the card was fixed at 3.3V. Those

could damage a 1.5V AGP slot, and it is because of that card,

that "AGP Protect" circuits were added to 845 motherboards

like the one you're working with. My P4B 845 motherboard

revision 1.05, was the first revision of that motherboard,

to be protected against SIS305 cards or similar (improperly

keyed cards). A red LED on the motherboard, checks TYPEDET,

and if no ground is present, that indicates a damaging

SIS305 could have been plugged in. The protection circuit,

prevents the ATX power supply from being turned on. And

the red LED glows as a hint as to how you screwed up.





3 ATI radeon 7000, and VE = Not work 100%
1 GeForce2 MX 400 = Work fine

wrote:
On Tuesday, September 17, 2013 7:48:13 AM UTC-7, Paul wrote:
wrote:

Was Intel MB designed not to support ATI Radeon model graphic cards?
I tried 3 Radeon cards on Intel D845PT P4 and none of them work perfectly.
The MB manual said "D845PT board is only compatible with 1.5 V AGP cards.
My card l think is all 1.5V 8x/4x/2x but I don't know for sure.
All 3 Radeon graphic work with AMD MB.
Anyone knows more detail on the incompatible issue between Intel P4 MB and ATI Radeon card?


3 ATI radeon 7000, and VE = Not work 100%
1 GeForce2 MX 400 = Work fine

http://www.playtool.com/pages/agpcom...iversalagpcard

"ATI Radeon VE Universal AGP Card
ATI Radeon 7000 Universal AGP Card

Universal AGP Card
Double slotted
Available speeds 1x, 2x at 3.3V
1x, 2x, 4x at 1.5V"

The cards physically, should have TYPEDET grounded.
So pin 2A should be connected to pin 5A with zero ohms.

"Intel 845 AGP 1.5V Motherboard

AGP 1.5V Motherboard
1.5V keyed
Supports only 1.5V signaling.
Available speeds 1x, 2x, 4x
"

So it should have worked, as long as the motherboard
delivered 1.5V to the slot.

*******

http://download.intel.com/support/mo...t/a7585201.pdf

"1.13.1 Hardware Monitor Component

Power supply monitoring (+5 V, +3.3 V, +1.5 V, 3.3 VSB, and Vccp)
to detect levels above or below acceptable values
"

Using a copy of Speedfan or MBM5, then that gives you a
way to verify the 1.5V level on the AGP slot. That is unusual,
in that a lot of other brands of motherboard, would not
have a connection to 1.5V and 3.3VSB. You're interested
in the 1.5V level. There is no guarantee Speedfan can
find the monitor chip, but give it a try.

http://www.almico.com/speedfan449.exe

Now, how you'd use that in practice, is with two video
cards at the same time. Plug a PCI video card into
a PCI slot. Connect the monitor to the PCI video card.
Set the BIOS video card detection to "PCI before AGP".
Insert the non-working AGP card as well. Start the
system. If the motherboard BIOS has a "hardware monitor"
page in the BIOS setup screen, you can verify the 1.5V
level in there. Then, shut down the system if you're
not happy with the results. I keep a single PCI card,
precisely for these reasons, for debugging AGP or PCI
Express cards that don't work. With the PCI video card
present, you have a means to work at the BIOS level or
boot the OS, then look at the hardware characteristics
of the (broken) cards.

That's about the only thing I can suggest in terms
of diagnosis.

Using Speedfan with the MX400 present, is bound to
show a properly working AGP voltage. But you can
still do a check if you wish. And since the card
works, you won't need a second video card to try it.
The voltage level could be different, with the MX400
plugged in, versus the ATI 7000. Or, it could be
the MX400 tolerates the situation better. Standards-wise,
the MX400 is the same as the other cards - Universal AGP Card.

Paul