CPW? vs System Performance

The discussion in the other thread about disk sizes, density, cache and
arms has been very interesting since we are facing the same issues.
Now I have an email from IBM that really confuses me (as if the whole
disk thing was not confusing enough)...

o  1st Generation 8xx Processor was 2X the speed of the 7xx Processors,
2nd Generation speed improved 20%, 3rd Generation more than 2X again.
      When Under 80% CPU Utilization, a 1st Gen.  8xx processor with the
same CPW as a 7xx processor, will run twice as fast as the 7xx.
            ie:  More Work with Better Response Times.
      8xx Fast Clock Speed & L2 Cache is required for CPU Intensive
Workloads, like Lotus Domino, WebSphere, Java, & e-Business.

Is this correct? How does something with the same CPW run twice as fast?
What is the effect
of CFINT in a server model? And is the 270 part of the solution or problem
(ie more 7xx than
8xx)?

The PC server and Unix server area seems so much less complicated than this
model.
Big drives, fast processors, lots of cache = performance.

So is CPW obsolete? I had always understood it was a measure of total system
throughput,
not just cpu capability.

Regards

Konrad Underkofler