[rescue] O2 graphics compared to entry-level Octane graphics?

[Kurt Huhn]

Bill Bradford wrote:
> 
> How do "onboard" O2 graphics compare to I2/Extreme and
> entry-level Octane (SI?) graphics?
> 

In my experience, and according to what I've gleaned from sources
online, The graphics capabilities go beyond just the framebuffer.  The
O2 has Unified Memory ARchitecture (UMA) and can use large sections of
physical (aka system) RAM as texture RAM.  The O2 RAM has a bandwidth of
about 2GBps, which allows for the high-speed needed to maintain textures
on the display.  It does have (AFAIK) hardware Z-buffer, but it does
*not* have any geometry engines in the graphics hardware.  It does not
(AFAIK) have the capacity to do lighting on the hardware.  Both high
geomoetry levels, and lighting, can slow down the display significantly
(using Lightwave as an example)

The Octane SI graphics have geometry processors, whith the ability to
add texture RAM modules (TRAM).  Without TRAM, the Octane graphics
appear anemic when manipulating textured surfaces (like morph in the
demos, or texcube).  However, for straight geometry, with lighting and
shadows, the Octane SI is very good  - if not excellent.  The SSI
basically doubles the number of processors on the graphics board, and
allows for a second TRAM module.  SE is a later revision and offers
roughly 30% speed increase of SI.  Adding TRAM modules to an SSI board
will make that board an MXI graphics system.  SI with TRAM is nice, but
there isn't a SGI marketing name for it other than SI+TRAM.

I mentioned Lightwave, wich I use on the Octane almost exclusively.  The
Octane is very good about displaying and manipulating models in 3D, on
the O2 it can seriously suck if you have a complex model.
-- 
Kurt
kurt at k-huhn.com