Free Republic
Browse · Search
News/Activism
Topics · Post Article

To: fork
I understand your loyalty. I first worked with Apple on the Lisa. But the reality is, any networked systems, be they Apple or Intel/AMD based can't match the raw power of a mainframe. The networked systems are a curiosity. Just do the math, how many bits can you shovel down a T1 line? Compare that to a Cray 1024 bit wide parallel bus running at say 500MHz. Now that is shoveling shi!t. And without being insulting when was the last time you worked with an interleaved memory system?
FWIW, the most powerful computers are not even on the list.
33 posted on 03/24/2005 8:49:41 PM PST by ProudVet77 (It's boogitty boogitty boogitty season!)
[ Post Reply | Private Reply | To 30 | View Replies ]


To: ProudVet77
Aye, theoretical top performance can be quite different from sustained performance on real-world problems.

Partitioning your data grid can be a pain, as well as the issues of communication latency.

Full Disclosure: NEC Earth Simulator is still vector-based.

Second disclosure: For "embarassingly parallel" or "near embarassingly parallel" (Monte Carlo, massive searches, among others) the Blue Gene type is the way to go. But there was just something so--classy--about the Cray chassis.

Cheers!

42 posted on 03/24/2005 9:10:26 PM PST by grey_whiskers (The opinions are solely those of the author and are subject to change without notice.)
[ Post Reply | Private Reply | To 33 | View Replies ]

Free Republic
Browse · Search
News/Activism
Topics · Post Article


FreeRepublic, LLC, PO BOX 9771, FRESNO, CA 93794
FreeRepublic.com is powered by software copyright 2000-2008 John Robinson