RE: How are CPU Speed and Overall CPW Related?

[Jim Damato <jdamato@xxxxxxxxxxxxxxxxx>]

Before I jump in headfirst on this one, please help me to understand what
we're talking about.  In your opinion, what do you think is the reason that
IBM is governing the CPU?  Is there a technical reason why they would want
to do it?  I'm a little unclear on the concepts that have been brought up on
this thread.

I've had a nagging feeling that IBM targets CPW performance in order to
arrange their product line.  Where Compaq (and I hate to make the
comparison) might get the new Intel chip, configure cache and memory, and
report the performance of various configurations, I wondered if IBM is more
cynical about it.  Once they determine the capabilities of the new chips and
architecture I felt that they designed target CPW ranges for the models.
After understanding the borderline-fraudulent practices that govern and
price Interactive Feature it wouldn't surprise me if they drag down the
processors on some models to keep them within their target CPW performance
range.

Is this what you mean, or are you talking about something far less
nefarious?

James Damato
Manager - Technical Administration
Dollar General Corporation
<mailto:jdamato@dollargeneral.com>


-----Original Message-----
From: Nathan M. Andelin [mailto:nathanma@haaga.com]
Sent: Wednesday, May 02, 2001 1:12 PM
To: MIDRANGE-L@midrange.com
Subject: Re: How are CPU Speed and Overall CPW Related?


I'm devoted to OS/400.  I build OS/400 based applications for people.  My
current focus is Web development.  I'd like my applications to be able to
compete against their Windows counterparts.  But my customers are faced with
a dilemma.  I'll try to illustrate with an example:

 D theString       S             50A
 D theShorter      S             50A   Inz('TEST STRING')
 D nbr             S             10I 0
 D Secs            S             10I 0

 D Now             S               Z
 D Then            S               Z

 C                   Time                    Then

 C                   Do        100000        nbr
 C                   Eval      theString = %trim(theShorter)
 C                   EndDo

 C                   Time                    Now

C     Now           Subdur    Then          Secs:*MS

C     Secs          Dsply

C                   SetOn                                        LR
C                   Return

This simple program displays the number of milliseconds required TRIM blanks
from a 50 byte string and assign the result to another string variable -
100,000 times.

Let's take a look at an equivalent program, written in Visual Foxpro, and
running under Windows:

theString = space(50)
theShorter = 'TEST STRING'
theShorter = theShorter + space(50 - len(theShorter))

then = seconds()

for i = 1 to 100000
  theString = alltrim(theShorter)
endfor

now = seconds()

? now - then

It takes 2095 MS to run the RPG code on my AS/400.  But it only takes 365 MS
to run the Foxpro code on my Laptop.  The Foxpro program offers 5.74 times
better performance.  But Foxpro is not faster than RPG.  The difference is
in the computer hardware.

My Laptop CPU offers a 330 Mhz clockspeed.  My AS/400 offers a 200 Mhz
clockspeed.  That explains a small part of Foxpro's superior performance.  I
believe the more meaningful difference is that AS/400 CPU is not fully
utilized.

Alexei Pytel gave me a hint that jogged my memory.  Some time ago, I read
that CPU throughput was governed largely by cache - not only clockspeed.  My
theory is that Rochester governs CPU throughput via processor cache.

A Midrange Computing chart shows that the same 200 Mhz Northstar processor
offers a CPW range of between 50 and 220.  Actually a 100 Mhz Apache
processor offers a CPW of 210.  This indicates that a lowly 100 Mhz
processor is capable of running my CPU bound RPG program 3 times faster than
my 200 Mhz processor.

Is "64 bit RISC" meaningful?  Is "silicon on wafer" meaningful?  Not if you
put a governor on the CPU.  Or, should I say, not if you withhold processor
cache?

Bottom line.  Web applications do large amounts of string handling.  And
that requires large amounts of CPU time.  Why not unbridle the AS/400 CPU?
Otherwise, Joe Foxpro, VB, Delphi, and C Sharp programmer is going to kill
me.

Nathan.


+---
| This is the Midrange System Mailing List!
| To submit a new message, send your mail to MIDRANGE-L@midrange.com.
| To subscribe to this list send email to MIDRANGE-L-SUB@midrange.com.
| To unsubscribe from this list send email to MIDRANGE-L-UNSUB@midrange.com.
| Questions should be directed to the list owner/operator:
david@midrange.com
+---
+---
| This is the Midrange System Mailing List!
| To submit a new message, send your mail to MIDRANGE-L@midrange.com.
| To subscribe to this list send email to MIDRANGE-L-SUB@midrange.com.
| To unsubscribe from this list send email to MIDRANGE-L-UNSUB@midrange.com.
| Questions should be directed to the list owner/operator: david@midrange.com
+---