Bill,
<The real question for SGI is can they keep ahead of the Wintel price/performance curve.>
Yes to a certain extent.
But, I don't agree with:
<Certainly there will always be a money-is-no-object niche for those who require top performance regardless of cost but that sector will become an ever-shrinking percentage of the total pie over time.>
While this can happen, what SGI needs to be doing is expanding the pie! As has been mentioned before on this thread, becoming a DataWarehouse box, pushing the envelope on what can be done with 'older', i.e. cheaper mass but still specialize technology, like the N64. Even your example of the VR demo. As long as they push as hard and improve as fast, as say Microsoft did, when they first bought out a crappy program call Windows 1.0.
Forbes, in its April 7 issue, has an article about real-life simulations (how's that for an oxymoron <g>). Coopers & Lybrands is doing simulations to determine how best to market music sales to 50000 "imaginary, fad-following record buyers", with all assumptions based on "real data". It executes a trillion MIs for the one simulation on a 30K line program. Takes two hours. And runs on a $10000 Indy workstation from Silicon Graphics.
Mat |
