[comp.windows.ms] IBM's XGA and the 386SX

bert@helix.nih.gov (Bert Tyler) (01/31/91)

> >It requires a 386SX or better processor not only because it has a
> > few 386-specific instructions during POST, but because it maps
> > graphics WAY up high in memory.
> 
> I am not completely sure whether this is correct, but I always thought
> that the 386SX cannot address 4 GigaBytes of physical memory but only
> 16 Megabytes due to a 24-bit address bus. If this is true, I do not
> see how the 386SX would make a difference to the 286 in its ability
> to access memory that high.

You are both correct.  The XGA maps its video memory to one of eight
4MB sections of address space, all of which are above the 16MB "real
address space" limit of the 386SX, and that means that a 386SX cannot
access that 4MB linear address space.  All is not lost, however - the
XGA memory can *also* be accessed via 1 1MB aperture that is located
somewhere in the 1MB-16MB physical address space, which is an area the
386SX *can* reach.  Using this 1MB aperture would certainly be possible
using a 286 as well, but it's a lot *easier* using a 386 or 386SX, since
those chipsets can access that aperture as one big linear address space. 

Oh - you say that you find it easier to access your video memory
"SuperVGA" style, using the A0000-AFFFf segment and bank-switching?
Nooooo problem - the XGA video memory can also be accessed *that*
way.  In fact, that's how my programs access it, because it was
easiest for me to add the XGA to the list of "SuperVGA" adapters
we support using that scheme, and occasional bank-switching doesn't
materially affect the performance of our programs any.  Could be
a *big* difference to other programs, though.

> Next real question, is any one out there using it?

I have one, and am having a ball playing with it.  Of course, us
guys with PS/2 equipment don't have a lot of video adapters to choose
from <grin>.