SPECfp92

[Bill Meine [Sun Denver MTS]]

> What would be the most useful approach for you as an 
> end user evaluating computer hardware for purchase?   What changes, if any,
> would make SPEC results more usefully relevant to you?    Do any of the
> foregoing ideas appeal to you - if not, what should be done instead?

As you notice, the prospects for coming up with ever larger and more
complicated benchmarks that truely represent customer loads will never
hit the mark.  However, I don't believe that makes the smaller, more
isolated tests any less useful.  The real problem is that most
customers don't have the understanding or methodology to composite the
results from the "micro" benchmarks to arrive at some figure-of-merit
for their own environment.  SPEC has a compositing method but most
customers don't know how to select or weight the component benchmarks
and I don't know if anyone understands how to bring in the database,
I/O, etc. benchmarks.  It's what they need (especially when they don't
have the application built yet) if someone could just tell them how to
get it done.

-Bill