View Single Post
  #2  
Old December 5th 03, 08:36 PM
Paul
external usenet poster
 
Posts: n/a
Default

In article ,
(Oded S.) wrote:

The Asus P4s333 motherboard has a serious problem when it comes to
graphic cards that has agp x8.
you can all read the problems mentioned in the link below:


http://www.abxzone.com/forums/showth...threadid=54274

I personally tried 5 different cards and NONE worked! This is totaly
insane. The cards we

Gainward geforce fx 5200 128 ddr agp x8
Gainward geforce4 128 ddr agp x8
ST Lab geforce 5200 fx
Palit Daytona geforce 5200 128 ddr 64-bit agp x8
Palit daytona geforce4 128 ddr agp x8

Asus claims the problem is with the low power supply from the psu
(mine was 300watts), but other users, as you can see from the link
above, have power supplies of 400+ watts, and it still doesn't work!
the problem is the chipset. the SIS 645 cannot deal with these new
graphic cards, and no new BIOS version is in sight.

This is all they had to say about this issue:


Dear valued Customer,
Thanks for Contacting ASUSTeK.
in theory, the board can support this vga card. but because this vga
card is
high level, we think the power issue is the issue point. please
use one more powerful power supply to verify this issue.
Thanks for choosing our products very much.


ASUS, you totaly disappoint me!

Oded S.


I think the problem might be as follows:

1) I read that on motherboards that don't support 8X, the AGP speed
is read back at BIOS time as 2X instead of 8X.
2) In the P4S333 BIOS, you can see that the AGP setting has only
two values - 1X and 4X. Now, this is a big guess, but when an
8X card is plugged in, and shows up as a 2X card, you end up
with a mismatch between what the BIOS will support and what
the card appears to want. Kaboom.

The root of the problem may be that the AGP status register is
hard wired to 0 in this design, making it impossible for the 8X
capability to be read properly. After all, the chipset is rated
for 4X, so only has to comply with the wording of the AGP20
spec.

ftp://download.intel.com/technology/...oads/agp20.pdf (pg.251)
http://developer.intel.com/technolog...0_final_10.pdf (pg.48)

The real question would be why the SIS chipset cannot operate at
AGP 2X. Normally, backward compatibility requirements for AGP require
support for all rates, so a 4X mobo should be able to do 2X and 1X.
I expect this is some kind of limitation with the Northbridge chip
design. Maybe the mode is in there, but doesn't work properly.

The P4S533 manual shows all three options 1X,2X,4X in the BIOS, so
I guess the different version of Northbridge fixes it. I would be
curious about when a P4S333 is flashed with a P4S533 BIOS, whether
the AGP Capability entry in the BIOS offers 1X, 2X, 4X or not. If
it does, then the only thing preventing the hack from working is
whether the busted 2X AGP support just happens to work in that particular
Northbridge chip or not.

If you hack the slot to indicate a 3.3V only card, that would force
operation at 1X, but that would hardly be a good permanent fix,
because the AGP card performance would be pretty bad at 1X. That hack
would involve the TYPEDET pin, and assumes both the motherboard and
video card can operate safely with 3.3V AGP I/O.

Since the manual documents this deficiency early in the life of the
board (i.e. it is in the manual), I wonder why Asus bothered to ship
this board. They should have waited for the next version of that
Northbridge with a working 2X AGP rate.

My theories,
Paul