The idea that anyone can design, much less ship, an NT machine capable of even touching the performance of an Onyx^2, RealityMonster, or even an Origin2000 is absurd. The architecture of the chips, the bus, the OS, and the hard-wired graphics capability of these machines is untouchable. I know E&S (I have friends who work there), and they've been playing catch-up with SGI since SGI bought MIPS. Even if it's just a niche market, I don't believe NT can ever match the power of SGI/Cray. I'd guess that the announcement from E&S is just more vaporware product.
Do you have figures on how many PPro units have been shipped? I'd be interested to compare numbers with the new MIPS family (5K, 8K, 10K).
I was under the impression that the incompatibility with older software coupled with the fact that the PPro can't be used in laptops was negatively affecting orders. My main point was that sooner or later, instruction set changes in the chip will neccesitate a completely new suite of software. Think about it. If you had to buy all new software just to upgrade your processor, would you do it? What if better performance for the same money was available by migrating to another platform, would you do that? The point is that INTC is soon to hit the architectural limit on the x86 family, whereas that wall isn't even in sight yet for the MIPS based architecture. |
