🚀 go-pugleaf

Yuma,

I have to agree with many points in your post.  First of all, the test
to which Barbara Suplie referred was not a very scientific one.  The
deck was heavily stacked in favor of the Windows NT system (dual
processors with a lot more system RAM and video RAM than the Mac or Win
95 machine).  For it to be a fair test the Mac system should have had
dual CPUs with clock speed identical to the Pentium Pro opponent as well
as the same amount of RAM.

These two systems would have been better to use in the comparison:

Micron Millenia Pro with one Pentium Pro 200MHz CPU, 128MB EDO RAM,
Windows NT 4.0

UMAX S900L with one PowerPC 604e 200MHz CPU, 128MB of 128-bit
interleaved RAM, Mac OS 7.5.3

Both systems should be configured with similar hard drives, video cards
and other periphreals.

I suggest using single processor systems.  A dual processor match would
stack the deck in favor of the NT system since the Mac OS doesn't
support full symmetrical multiprocessing.  The Be Operating System would
be better for a dual CPU test since it has many of NT's bells and
whistles (support for multithreading, symmetrical multiprocessing and
OpenGL).

If we're talking about CPU performance alone, then here are the numbers
from the Intel, Motorola and Specbench web sites:  The Pentium Pro can
execute three instructions per clock cycle (600 million instructions per
second possible for a 200MHz Pentium Pro).  The PowerPC 604e can execute
four instructions per clock cycle (800 million instructions per second
possible for a 200MHz PowerPC 604e).  That makes the 604e 33% faster
than a Pentium Pro on how many MIPS it can do.  The floating point
performance of the Pentium Pro and PowerPC 604e are in the same
neighborhood, with the slight edge going to PowerPC.  The results of
SPECint95 and SPECfp95 from Specbench on both CPUs:

Pentium Pro 200MHz, 66MHz bus, 512K L2 Cache
SPECint95 = 8.58
SPECfp95 = 6.48

PowerPC 604e 200MHz, 66MHz bus, 512K L2 Cache
SPECint95 = 8.02
SPECint95 = 6.72

There are certain things that will slow both of these chips down.  Both
of these chips employ out of order execution, branch prediction and
register renaming.  That kind of stuff doesn't work well when fed
anything smaller than 32-bit code.  You'll get pipeline stalls.  This is
why Pentium Pros slow down in Windows 95 or Win 3.x.  This could be the
root of the Mac file system problem you mentioned.  When you say it
isn't PPC native, you are saying it isn't 32-bit right?

With these real numbers, anyone would have to agree that the PowerPC
604e is faster than a Pentium Pro of identical clock speed, system bus
and L2 cache.  Also, anyone would have to agree the PowerPC 604e is a
good bit faster, but not twice or three times faster.  That gets back to
that cost argument.  However, if someone wants the "cost is no object
speed demon" or your job turnaround time is very mission critical then
you really have to look at a DEC Alpha 21164 based system, or better yet
a Sun Solaris or Silicon Graphics MIPS R10000 based workstation.  Those
systems have jaw dropping performance numbers, and prices to go with
them.

I was pleased you brought up the point about system care.  When a user
doesn't maintain their system or bother to learn how to do so, they
don't have the right to complain when the system crashes.  Instead of
blaming the OS or CPU they need to blame themselves!  I know how to take
care of my Win95 machine, and because I do so on a regular basis I never
seem to have the problems so many other users complain about.

Best Regards,

Bobby