[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Multi-channel VGA card

From: Irv D Negrin

>>> >We intend to develop a PCI video card with two VGA chips on it. Have.
>>> >anybody had any experience in design such a card.
>>> >We are not sure what need to be done so that both VGA chips can be
>>> >accessible by Windows 98 and NT5.0 multi-monitors facilities.
>>> >Any advice or suggestions are welcome.
>>> One point from a hardware perspective: make sure you present only one
>>> electrical load to the PCI bus.  If there are two VGA chips, they must
>>> be behind a bridge.
>>  What is the reason for that requirement?  Assuming you can figure out a
>>  way to get the non-shared PCI signals (eg reqs, grants, interrupts,
>>  up to the card, wouldn't it be possible to extend the PCI bus so that
>>  one card could have more than one "virtual slot"?  Granted, it would
>>  probably not be considered a true blue "PCI card" but electrically it
>>  seems possible...
> Well, aside from the gross violation of a absolute rule in the PCI spec
> add on card shall present more than one load"), and the device
> architecture (each slot gets one and exactly one config space unless its
> bridge or a multifunction device), and the fact that it flat out wouldn't
> work, um, I see nothing wrong with it.
> You will HAVE to put the VGA's behind a bridge unless one VGA is custom
> designed to have a bridge function built in and the 2nd VGA is on this
> backdoor bridge or the one VGA chip has two device functions built in
(i.e. is
> two vga's in one)..

> I believe there are also some OS issues with initializing boot devices
> behind a bridge, but I gather these have been solved due to the existance
> various multi-VGA boards already on the market.

The bridge must be capable of claiming bus cycles to the default VGA memory
and IO addresses, passing them through to the VGA devices. It should also
support VGA palette snooping. This is necessary for bios to configure the
display using the legacy ISA VGA register and memory buffer addressing

Irv Negrin
Eastman Kodak Company