At 20:56 -0500 05/17/2005, Donald Lee wrote:

>FWIW - I'm almost certain that the 9500 did *not* have a PCI-PCI
>bridge on the mboard.  There were two bus controllers.  This is
>why the UMAX boxen were a little strange with some PCI cards, cuz
>Mac OS 7/8/9 didn't support the PCI-PCI bridge properly.

It wasn't just the OS that didn't support the bridge properly, the 
firmware in those machines is buggy with respect to PCI-PCI Bridges. 
Specifically, if you try to daisy chain bridges, i.e. install a PCI 
card bearing a PCI bridge in a PCI slot which is behind a PCI bridge, 
the machine will freeze during initialization (grey screen).

There are some details to that.  If the bridge bearing card is the 
only PCI card behind the first bridge, then the machine will work 
okay.  If there are other cards behind the bridge, but they are cards 
with no firmware on board (ethernet, USB, firewire) then the machine 
seems to work okay.   But do something like putting an Adaptec 
3940U(W) in a lower slot of the Umax S900 (the 3940U and UW have a 
DEC 21052 on board) and also put a video card, SCSI card, or ATA card 
in another lower (behind the bridge) slot and the machine will freeze 
before it can begin loading any OS.

I've tested this behaviour with all three revisions of the firmware, 
$77D.28F2 (original X500 and clones), $77D.34F1 (first model 8600 and 
9600 and 7300) and $77D.34F5 (revised (Kansas) 8600 and 9600).

The 9500 does have two CPU bus to PCI bus bridges/controllers, and so 
is different from the S900.  I.e., the 9500 has two Bandit chips. 
The S900 has one Bandit chip and replaces the third PCI slot with a 
PCI-PCI Bridge to provide four additional slots.

Jeff Walther