🚀 go-pugleaf

Arun Gupta (gupta@mrspock.mt.att.com) wrote:
: On Byte benchmarks, my 7100/66 ( 601 PPC at 66 MHz, main bus 33 MHz)
: runs at between 0.90 and 0.95 on integer benchmarks depending on
: the level of optimization, on which the Dell XPS P-90 (Pentium 90 MHz, 
: main bus 66 MHz) by definition scores 1.0.  On floating point
: multiplication and division my PPC scores 1.2 .  On floating point
: using the exp(), sin(), cos() functions, my PPC however, is 10-15
: times slower than the Pentium.  I am told this is because of a rather
: poor math library -- I will find out more about this as I update my
: system software from the now-venerable System 7.1.2.

I think that this may be because these cosines stuff is supported in hardware 
and not by a library on an X86 chip. 

: The compiler used was CodeWarrior 7.0.  For a comparison between
: CodeWarrior 7.0 and the Motorola compiler, which generates faster
: code, visit the PowerPC pages on the Motorola site.  

And the Dell XPS P-90, by definition, the 1.0 standard, which compiler did it
use? could it not use a better compiler?

: The benchmark results are astonishing : would you have expected 
: a 66/33 MHz system to be remotely comparable to a 90/66 MHz system ?

: There are several things, IMO, masking the true performance of the
: PPC systems.  One is a technological lag.  The fastest PPC system
: I'm aware of has a 50 MHz bus, while the Pentiums are at 66 MHz.
: PPC systems don't use EDO RAM.  The most common PPC OS - MacOS is
: not fully native.  Even in chip technology, the PPCs are at 0.5 micron,
: while the Pentiums are moving to 0.35 and 0.25 micron.  


PPC601-100 is at .35 micron.  
PPC604-100,120,132 is at .5 micron
PPC604-150 is at .35 micron.

P5-100 is at .6 micron
P5-133,150,166 is at .35 micron. 
P6-150 is at .6 micron
P6-200 is at .35 micron

production quality .25 micron process still has not arrived. Maybe in '97 or 
'98?  

: PPC systems are able to give competitive performance to Pentiums
: with slower system parts.  The PPC chips are price competitive to 
: Pentiums.   Why then haven't PPC systems clobbered Pentium systems
: on price ?

: The main reason, IMO, is that Intel systems have 10-20 times the 
: volume of PPC in sales, and so costs are likewise recovered over
: many more systems.  Note that if Apple abandoned backward compatibility
: and produced a brand-new OS, the PPC would have a marketshare of
: zero.  The lousy math library that I have for my PPC is illustrative
: of the difficulty involved in exploiting the new architecture.

: The limits on the PPC are economic, and not technological.
: As PPCs move up the technological curve, Intel will find itself
: running harder and harder to stay competitive with the x86 
: architecture.

Chicken and Egg problem. No one is able to over take Intel on the technology
front until they invest the money to get and edge on processor architecture/
processor design/process technology/fab capacity.  But if they don't already
have the edge, what can make people purchase their systems, so that they
can get the money to invest? And no, Intel won't find a hardtime staying
competative with its own X86 architecture.

: -arun gupta