From owner-freebsd-smp  Fri Oct 25 23:08:44 1996
Return-Path: owner-smp
Received: (from root@localhost)
          by freefall.freebsd.org (8.7.5/8.7.3) id XAA01449
          for smp-outgoing; Fri, 25 Oct 1996 23:08:44 -0700 (PDT)
Received: from GndRsh.aac.dev.com (GndRsh.aac.dev.com [198.145.92.241])
          by freefall.freebsd.org (8.7.5/8.7.3) with ESMTP id XAA01444;
          Fri, 25 Oct 1996 23:08:40 -0700 (PDT)
Received: (from rgrimes@localhost) by GndRsh.aac.dev.com (8.7.5/8.7.3) id XAA11759; Fri, 25 Oct 1996 23:08:06 -0700 (PDT)
From: "Rodney W. Grimes" <rgrimes@GndRsh.aac.dev.com>
Message-Id: <199610260608.XAA11759@GndRsh.aac.dev.com>
Subject: Re: Recommendations...
In-Reply-To: <199610252255.PAA29409@sunrise.cs.berkeley.edu> from Satoshi Asami at "Oct 25, 96 03:55:08 pm"
To: asami@FreeBSD.org (Satoshi Asami)
Date: Fri, 25 Oct 1996 23:08:06 -0700 (PDT)
Cc: ejs@bfd.com, michaelv@MindBender.serv.net, scrappy@ki.net,
        current@FreeBSD.org, smp@FreeBSD.org
X-Mailer: ELM [version 2.4ME+ PL25 (25)]
MIME-Version: 1.0
Content-Type: text/plain; charset=US-ASCII
Content-Transfer-Encoding: 7bit
Sender: owner-smp@FreeBSD.org
X-Loop: FreeBSD.org
Precedence: bulk

>  * What low memory bandwidth on the Natoma???  That thing smokes when comparied
>  * to a 430HX chipset.
> 
> That contradicts our findings.  A P5-133 with Triton or Triton II can
> move 70-80MB/s (depending on EDO or non-EDO), but I can't get more
> than 45MB/s out of a P6-200 with Natoma/server (at least that's what
> Intel told us).
> 
> Satoshi
> 
> P.S. Details on "http://now.cs.berkeley.edu/Td/bcopy.html".

Okay, finally, the routes came back to life.  Your ``P6/Natoma/Intel/server''
was probably a pre production pile of shit.  I have a P6 system just getting
ready to go to burn in with a 4G quantum Atlas.  If you have the time
tomorrow I will set it up outside the firewall and let you run benchmarks
on it all day long. 

I think you will find that a current AAI production Pentium PRO 200 system
running on the ASUS PCI/I-P6NP5 is quite a blinding fast machine.  In this
configuration it turns make worlds in about 1:38 minutes, and using
multiple spindles it has been seen to go as low as 1:22.  All times are
2.1.5 stable with the following /etc/make.conf tweaks:
    CFLAGS= -O -pipe
    NOPROFILE=      true
    NOMANCOMPRESS=  true
    SHARED=         copies

My business has me far to overloaded to run benchmarks for you, but
your welcome to come on over and play with the box. 

I can turn parity off and get another 8 to 12% out of the memory
system.... or turn ECC on and slow it down 8 to 12%.  Since those
require BIOS tweaking it can be arranged for me to do a couple of
reboots with tweaks.

-- 
Rod Grimes                                      rgrimes@gndrsh.aac.dev.com
Accurate Automation, Inc.                   Reliable computers for FreeBSD