🚀 go-pugleaf

 > I am attempting to quantify the difference in hardware costs to outfit 
 > administrative assistants with PCs/Macs of `equal' power.
 > 
 > Is it reasonable to assume that I can compare Motorola & Intel driven 
 > systems on the usual sorts of things - MHz, cache size, w/wo FPU, bus 
 > widths, etc?

No. You cannot compare across different CPU lines and different computer
architectures by using simplistic measures like Mhz, cache size, etc.
You can't say that a PC compatible using a 386 CPU at 33Mhz is "clearly"
twice as fast as a Macintosh LC-II at 16Mhz, because that is not
necessarily the case. How much benefit you get from cache sizes even
within the same CPU line depends greatly on the operating system and the
kind of applications you are running, and PC's and Macintoshes are
disparate in those respects. I suppose the presence and lack of FPU's is
something you can consider to make equivalent amount of difference,
relatively speaking, for both PC's and Macintoshes, but that doesn't
mean that you'll obtain the same amount of performance boost by adding
an FPU in both cases. And with respect to the expansion bus, comparing
ISA, EISA, MCA, and PCI- and VL-local buses on the PC side with Nubus,
SCSI, and Direct Processor Expansion slots on Macintoshes would be
comparing apples and oranges, for the most part.

Think about what you are trying to do. I won't try to argue that
"administrative" assistants don't need powerful machines, but they, like
most computer users, are interested in how much productive throughput
they get from the machine for the kind of software that they run and not
the raw specs (yeah, the raw specs are easier to compare and get --
speaking of which, I've never seen any "SPECmarks" published for Macs
for A/UX). Unfortunately, there are no studies I know of which try to
document how much of a difference the "speed" of your machine makes in
your productivity, given the same product line and software/OS (e.g. it
might be the case that for Macintoshes, doubling the raw speed would
nearly double user throughput, but on PC's doubling raw speed would only
improve user throughput by 1.5, *OR* vice versa). This information would
allow you to understand how much value you get for your money as you get
spluftier and spluftier models within each line, and then you would have
to establish some cross-platform baseline.

It also seems to me that extensive discussion of this issue doesn't
really belong in this newsgroup, but I don't know where to redirect.

--
| Ping Huang (INTERNET: pshuang@martigny.ai.mit.edu) speaking for himself.
| "One Thing to name them all, One Thing to define them,
|  One Thing to place them in environments and bind them..."