Display adapter for X11R3 386/ix (cont.)

[Scott Wiesner]

>From article <1990Mar28.032151.7871 at ddsw1.MCS.COM>, by karl at ddsw1.MCS.COM (Karl Denninger):
> 
> Next, buy a board which has 
> an on-board PROCESSOR.  That means something like the Matrox board, or 
> the Pixelworks board, etc.  I'm not sure about the 8514/A, it may have one.
> You especially want area-fill and line-drawing intristics on the board;

The 8514 has lines drawing, rectangle filling, and blitting in hardware.
It's very fast at moving windows, scrolling, text painting, etc.  It's 
one of the fastest boards we've seen.

> .... Lots of VGA discussion deleted...

Yes, the VGA is a pretty crufty device.  We've spent a fair amount of time
trying to wring some reasonable performance out of it.  The thing that's 
most noticably slow is the general blitting (window moving).  Character
output, line drawing, and to a lesser extent area fill aren't too bad.  If
you want to speed up the window moving and don't mind sacrificing colors, 
you can specify 2 or 4 colors instead of 16 in the Xconfig file.  This
helps out quite a bit for moving windows around because there's less data
to move.

I've seen a fairly large difference between the various VGA cards.  Older
cards, such as the old STB, the Orchid ProDesigner, etc, are only fair.
Newer ones like the Video 7, Paradise, and upcoming ones from STB and
Orchid are noticably better.  The issue of 8 and 16 bit cards is just
noise as long as you're talking about 16 color VGA support.  16 bit cards
show their worth when you move to a 256 color mode.  It makes a big
difference here, with 16 bit cards running up to 2 times as fast as
8 bit cards.  How do I know about this?  Well, you see, we've got a
new server coming that will support 256 colors.  It's not as fast as
the 16 color server, but it's reasonable, and for people who just have
to have more colors, it's a low cost alternative.

> "X11R3" isn't known as being especially efficient either... R4 is supposed
> to be much better, but I haven't had the chance to play with it yet (there's
> this little thing called a server that isn't there yet for 386 displays......)

Much of the MIT R3 slowness was due to a lack of specific code for drawing
on color displays.  R4 added new code to rectify this, but of course, we
had already done this for VGA support, so you won't see the kind of dramatic
improvement that is shown on something like a Sun between R3 and R4.

Scott Wiesner
Interactive Systems
X Development Group