30 May
2011
30 May
'11
9:34 a.m.
On 5/30/11 11:27 AM, Chuck Guzis wrote:
The big enemy of supercomputer engineering today is
Moore's Law. By
the time design and testing is complete and product is rolling off
the line, it's obsolete. Better to spend your time building a fast
bit of silicon and hoping that it'll see a 6 month to one year
product window.
As far as semiconductor process evolution goes, definitely. The type
of innovation that can make an impact and last a while is architectural.
Going through the effort of designing an building a
mainframe-style
supercomputer such as a Cray 3 would be folly today. Witness Steve
Chen's "box full of Pentiums" as how badly this sort of thing can go.
Pentia? ;)
-Dave
--
Dave McGuire
Port Charlotte, FL