Steve Gifford (sjg++at++terrex.com)
Mon, 30 Aug 1999 18:20:10 -0700
This is quite true. In my experience (with PC based systems in
particular), disk bandwidth is the limiting factor, followed by texture
fill rate, with polygon throughput a very distant third. Particularly for
geospecific databases you spend most of your time trying to get that data
into texture memory without negatively affecting your performance.
However, it's surprising what you can do with a well structured terrain
database.
>It seems more than a little optimistic to say a PC can just do this.
>Sure it can do something but it can't handle the same class of problem.
>You have to compare apples with apples, doing otherwise is exactly the
>kind of loosely defined approach to analysis which has led to problems
>of programs like CCTT where they realized too late that PC's weren't
>going to do the job.
To draw this conclusion you're setting a very high bar for the definition
of a "geospecific database." This is interesting, I think, for historical
reasons primarily. Evans & Sutherland used to (and I believe still) do the
same thing with with the term "real time" against SGI in sales situations.
They lost market share anyway. I think there may be a lesson there. :-)
On a more serious note, I will say that we have customers generating
geospecific databases and deploying their applications on Windows boxes
(some of them sold by SGI) with no complaint from *their* customers. Since
they are (obviously) not doing so with Performer I think more discussion
along those lines is probably not appropriate to this forum.
-Steve Gifford
TERREX Inc.
This archive was generated by hypermail 2.0b2 on Mon Aug 30 1999 - 18:24:06 PDT