lipman@dtrc.dt.navy.mil (Lipman) (11/29/90)
We've got $150,000 to spend on a computer graphics visualization system. We want to look at results from CFD and finite element analyses. Currently, we have an Apollo network with about 25 nodes. What would you buy and how would you configure it? SGI, IBM, SUN, STARDENT? What about networking to Apollo? What about benchmark speeds for real applications? Thanks in advance, e-mail please Bob Lipman
eachus@linus.mitre.org (Robert I. Eachus) (12/09/90)
This question has a two-part answer... If you are interested
primarily in one person playing with data, you need the fastest
graphics horse for your application which may be an IBM RS/6000 or an
IRIS.
However, assuming that you are simulating non-static systems,
like fluid flows, then if you are concerned at all with "publishing"
the results of your work, then a system which can produce video tapes
and/or run a projection display is a must. "Everybody gather round"
type dog-and-pony shows have lots of disadvantages. Among them are
that you can't send a tape to a sponsor or to show your results at a
meeting on a projection TV system.
Almost everyone who does this sort of stuff professionally now
uses Amigas. Now that the Video Toaster is shipping, the answer as
to what sort of Amiga setup to get is almost forced:
Amiga 2000HD with A225 Ethernet Board
GVP 3001 board (your choice 25,33, or 50 MHz) with 8+ Meg of memory
NewTek Video Toaster
Two SVHS VCR's at least one with single-step capability
Three high resolution monitors to taste
If you add a couple of good camcorders and synthesizer (I can
imagine trying to get that through the Navy!) you would have a
complete TV production studio. But for what you need the total system
cost should be about 15K. Throw in some Amiga 3000s for preliminary
work, ray tracing, and the like, and there is still plenty of money
for a high-end graphics workstation for exploratory modelling. (Not
that this system will be a dog. Exactly the opposite! Six to 13 MIPS
with probably the fastest graphics display accelerator available
anywhere (the Toaster). It's just that this setup--and its
controls--are aimed at video production not exploration. You often can
do a much better job of exploratory data analysis on some of the Iris
workstations for example.)
The reason for choosing the GVP 68030 board is that it permits
more high speed memory than the A2630. The Video Toaster can be
forced into an Amiga 3000. But until NewTek updates the board it is
not recommended, and you will probably need the extra slots in the
2000 chassis anyway. Assuming that you will be using this system with
a high-powered Unix file server, you will want the color version of
Dale Luck's X-windows software for the Amiga. Just being able to do a
good job of recording X displays directly on video tape is more than
enough to justify it. Throw in some paint and 3D programs and
software will add about 2K to the system price.
--
Robert I. Eachus
with STANDARD_DISCLAIMER;
use STANDARD_DISCLAIMER;
function MESSAGE (TEXT: in CLEVER_IDEAS) return BETTER_IDEAS is...