Wirt wrote:
>What a difference a week can make. Today, IBM broke even those
bounds. It
>announced plans to start work on a 10,000 MHz processor.
========================================================
Wirt, the article you excerpted said IBM had contracted with DOE to build
a supercomputer "capable of executing 10 trillion instructions per
second".  This is definitely NOT the same thing as a 10,000 MHz
processor!  First of all, a machine cycle is undoubtedly longer than one
clock pulse.  Secondly, we don't know whether the processor under
development will complete one instruction per cycle.  The third salient
point is that it is extremely doubtful that the supercomputer being
developed is a single-processor machine.

These are some of the same reasons that clock speed is a virtually
useless number for comparing the performance of two machines of
different architecture.  A case in point: HP9000's (and 3000's) frequently
outperform machines with much faster clocks when compared on the
basis of transaction rates.