delorie.com/archives/browse.cgi   search  
Mail Archives: djgpp/1996/07/03/10:16:15

From: kagel AT quasar DOT bloomberg DOT com
Date: Wed, 3 Jul 1996 10:07:02 -0400
Message-Id: <9607031407.AA06089@quasar.bloomberg.com >
To: orly AT gibson DOT eee DOT upd DOT edu DOT ph
Cc: eurgain AT enterprise DOT net, djgpp AT delorie DOT com
In-Reply-To: <Pine.SGI.3.93.960703132633.3942A-100000@gibson.eee.upd.edu.ph> (message from Orlando Andico on Wed, 3 Jul 1996 13:31:31 +0800 (GMT))
Subject: Re: hello world compile time
Reply-To: kagel AT dg1 DOT bloomberg DOT com

   From: Orlando Andico <orly AT gibson DOT eee DOT upd DOT edu DOT ph>

   On Tue, 2 Jul 1996, Alistair Hamilton wrote:

   > >Even Win95 does a reasonable job, about 6 secs the first time, <3 secs
   > >for subsequent compiles on a P75 with 40M RAM.
   >                                                 
   > 40 Meg???? Yegods! How the other half live!

   Yegods! is right! and on a P75! (suffer me to use a 32MB SGI!) on that
   score (since a P75 isn't exactly a speed demon...) what gives a greater
   overall performance improvement, tons and tons of RAM, or a faster
   processor? the biggest PC box I've ever had was a 16MB P133.. although I
   have seen someone running VHDL simulations on a 64MB i486 (!) doesn't
   anyone find that obscene? or does mucho memory really help more than CPU
   muscle?

Good lord!  A P75 is "not exactly a speed demon..."?  I'm still working on an
AMD 386/40 with 8MB at home!  Though I alternate between a P100 w/40MB and a DG
Aviion 9500 with 24 CPUS and 3GB RAM (no typo) at work so imagine my
frustration at home!  I first programmed PC's on a 8MHZ 286 (original IBM AT
generation 2)!  To me P75's fly!!!  (And I moved to PCs from a 2MHZ DEC20
supporting 25 programmers and 200 users so that AT was a dream.)  Boy are you
guys spoiled.

As to whether memory helps more than a faster CPU that depends on what you are
doing.  DJGPP, and especially ld, will use as much virtual memory as needed to
compile and link.  With larger C and especially C++ programs this can be quite
a bit.  More memory means less swapping and since disk is MUCH slower than CPU
(and yes I started on IBM 80ms access time disks, my first 31ms disk made me
feel like I had my own Cray) even at 8ms seek and 7200RPM with a 20MB/sec
transfer rate, memory CAN make a bigger difference than MHZ or processor class.
Of course upgrading from 8MB to 16 MB will gain you more than moving from 16MB
to 32 MB which is more effective than upgrading from 32MB to 40 or 48MB so the
law of diminishing returns definitely applies here.

Applications that are disk bound will not benefit at all from more RAM and only
slightly from faster CPU (due to faster DMA behavior from a faster clock and
faster buffer copies by the CPU).  Only CPU bound applications benefit
proportionally when the CPU is upgraded.  Also applications doing heavy
graphics will benefit from a faster clock if the video board in on a CPU bus
(PCI or VESA bus) card.

-- Art S. Kagel, kagel AT quasar DOT bloomberg DOT com

A proverb is no proverb to you 'till life has illustrated it.  -- John Keats

- Raw text -


  webmaster     delorie software   privacy  
  Copyright © 2019   by DJ Delorie     Updated Jul 2019