🚀 go-pugleaf

Garby Leon (garb@ix.netcom.com) wrote:
: 
: The 200Mhz Pentium Pro Spec benchmarks (8.09 SPECint95) show that on 32
: bit code, it's integer performance beats anything now available,
: including the 300Mhz Dec Alpha (7.33 Specint95) and 133Mhz Power PC 604
: (4.07 SPECint95).

Well, besides the fact that the 604 is available at 150 MHz, you're 
comparing "preliminary SPECs" versus SPECmarks in a selling computer.

We won't even go into the troubles of comparing SPECmarks on certain 
machines, which can be a dead end- the new UltraSPARC machine gets beat 
on SPECs by the PPro 150, but due to the architecture of the Ultra 
machines, they *murder* the PPro in real operations (like running 
operations on graphics arrays).

Intel always releases these incredibly impressive numbers a few months 
before they make a chip ready for market, and then when people start 
making computers with that chip, everyone's really surprised to find that 
the *real* machines end up running anywhere from 20% to 30% below those 
original numbers.  And real-world performance is always a *lot* slower 
than even the honest numbers.  And then Intel finds "bugs" in their 
compilers, bringing the numbers down some, or just quietly change the 
numbers on their website, rewriting history...

: It even beats the estimated performance of any Power
: PC chip yet talked about including the new PowerPC 620 (225 Specin92).

The 620?  How about the 604e?  The current line on the 620 is that it's 
being redesigned, and the last I heard, the 604e should be out sometime 
this year in a faster (200 MHz+) model that should give performance about 
even with the real-world PPro 200 MHz machines.  You know, about the same 
time they make PPro computers that really *use* the 200 MHz chip (not just 
old Pentium mother boards that have been hacked to hold the PPro).  
And about six months before they make PPro computers that have anything 
*near* the currently-announced speeds.

And you know the cool part?  Most of the mid-to-high range PMacs will be 
easily upgradable to the faster chip, due to the daughterboard that's 
used on many of the newer machines.  So the 7500 I'm planning on getting 
this month (or possibly a Power Computing PowerCurve model) could soon be 
a 200 MHz 604e...

: Too add insult to injury, some industry analysts that have taken an
: in-depth look at the architecture and are speculating that the
: architecture will support clock frequencies nearly twice as high as
: those released. It may be possible that Intel is sandbagging on Pentium
: Pro performance in order to keep sales of Pentiums strong.

...or it's possible that the analysts are just a little bit wrong.

Not to mention that little company in California that is working on a 700 
(yes, seven *hundred*) MHz version of the PPC. 

: You have to be quick to keep up with the Microprocessor market.
: The Byte benchmarks and the PC magazine benchmarks are last years news.
: For more information check out the performance section of Intels Web
: site at:
: 
:    http://www-us-east.intel.com/procs/perf/index.html

You also have to keep up with the news.  I wouldn't trust Intel's numbers 
at all.  They were "estimated" using Intel "reference computers" running 
software generated on Intel's "reference compiler," and have no bearing 
on reality. other than as an example of how far someone can stretch the 
truth in order to gain sales.

-- 

Chad Irby           /    My greatest fear: that future generations will,
cirby@gate.net     /      for some reason, refer to me as an "optimist."