[Prev][Next][Index][Thread]
good bang for buck machines?
-
To: info-dylan@ai.mit.edu
-
Subject: good bang for buck machines?
-
From: brucehoult@pobox.com (Bruce Hoult)
-
Date: Sat, 20 May 2000 17:45:02 -0400 (EDT)
-
Organization: The Internet Group Ltd
-
Xref: traf.lcs.mit.edu comp.lang.dylan:12041
Hi fellow Dylan hackers,
I'm looking at replacing my trusty Pentium Pro 200 linux machine with
something a bit more modern. It still goes pretty well, but finding 72
pin ECC RAM these days is an expensive pain, it's got a proprietry
motherboard design so I can't just swap that out etc etc (a lesson I've
learned for my 2nd x86 purchase). And now that Apple doesn't use SCSI
drives any more their price has headed for the stratosphere relative to
IDE which makes having an all-SCSI machine a dubious proposition when
you've got enough RAM that you don't do any real disk I/O anyway...
The prime performance driver for the new machine -- the only thing I'm
likely to do with it that will be slower than instantaneous -- is doing
program builds and the slowest of those are working on/with d2c. So I'm
really interested to hear from people with experience with current model
x86 chips. In particular, I'm interested to hear of any significant
difference between PIII and Athlon performance on Dylan code.
At the moment I'm leaning towards the 700 MHz Athlon on an Abit KA7 board
with 192 MB of PC133 RAM and some humungous IDE drive. The
Windows-oriented sites say that this is "da bomb" compared to earlier Asus
K7M or FIC SD11 boards that were apparently "da bomb" six months ago (and
can be found cheap now). The PC133 in particular apparently really
helps. When you compare a system based on this stuff with an
as-cheap-as-you-can-get Celeron or K6 system it's basically $250 - $300
more on a total system cost of $1000, but a lot faster.
One problem I'm having is finding realistic performance comparisions
between current systems and ones that are a couple of years old. You can
find all the comparisons you want between the K7 and the PIII, with
nicely-exaggerated graphs with a non-zero baseline, but how do they
*really* compare to a P166 or PPro200 or even a PII/300? And I don't mean
on Winstone.
So, if anyone out there is using d2c on a PIII or Athlon then I'd be
really really interested in hearing from you about what your machine
configuration is and how long it takes for, say, a "cd src/d2c;make
clean;time make" and/or the same thing in the src directory itself (which
runs all the tests, builds mindy and the demos etc).
Thanks!
-- Bruce
Follow-Ups: