> 
> I'm just guessing here, but I would think a "pcivga" driver would be
> needed that was capable of mapping a "VGA" card at non-VGA IO addresses.
> The current code in /sys/dev/pci/vga_pci.c just calls the code in
> /sys/dev/ic/vga.c which makes assumptions about the IO address space, etc.
> 
> What might be interesting is the output of "scanpci -v" showing
> those two cards.

pci bus 0x0 cardnum 0x08 function 0x0000: vendor 0x5333 device 0x8811
 S3 Trio32/64
 CardVendor 0x0000 card 0x0000
  STATUS    0x0200  COMMAND 0x0003
  CLASS     0x03 0x00 0x00  REVISION 0x54
  BASE0     0xe0000000  addr 0xe0000000  MEM
  MAX_LAT   0x00  MIN_GNT 0x00  INT_PIN 0x01  INT_LINE 0xff

pci bus 0x0 cardnum 0x09 function 0x0000: vendor 0x5333 device 0x8901
 S3 Trio64V2/DX or /GX
 CardVendor 0x5333 card 0x8901
  STATUS    0x0200  COMMAND 0x0000
  CLASS     0x03 0x00 0x00  REVISION 0x16
  BASE0     0xe4000000  addr 0xe4000000  MEM
  MAX_LAT   0x00  MIN_GNT 0x00  INT_PIN 0x01  INT_LINE 0xff

> Probably a card specific driver would be needed to communicate with the
> card in native PCI mode.


all i need is possibility of running two X servers (assuming i have 2
keyboards etc..).