From owner-freebsd-hackers  Mon May 19 16:24:00 1997
Return-Path: <owner-hackers>
Received: (from root@localhost)
          by hub.freebsd.org (8.8.5/8.8.5) id QAA25575
          for hackers-outgoing; Mon, 19 May 1997 16:24:00 -0700 (PDT)
Received: from mail.crl.com (mail.crl.com [165.113.1.22])
          by hub.freebsd.org (8.8.5/8.8.5) with SMTP id QAA25531
          for <hackers@FreeBSD.ORG>; Mon, 19 May 1997 16:23:27 -0700 (PDT)
Received: from conceptual.com ([209.60.202.194]) by mail.crl.com with SMTP id AA19132
  (5.65c/IDA-1.5 for <hackers@FreeBSD.ORG>); Mon, 19 May 1997 16:20:44 -0700
Received: from kofa (kofa.100mb.conceptual.com [10.0.1.3]) by conceptual.com (8.8.5/8.6.9) with ESMTP id QAA01644; Mon, 19 May 1997 16:23:16 -0700 (MST)
Message-Id: <3380E0D6.EB2A5EC7@consys.com>
Date: Mon, 19 May 1997 16:23:02 -0700
From: "Russell L. Carter" <rcarter@consys.com>
Organization: Conceptual Systems & Software
X-Mailer: Mozilla 4.0b4 [en] (WinNT; I)
Mime-Version: 1.0
To: sthaug@nethelp.no
Cc: hutton@isi.edu, hackers@FreeBSD.ORG
Subject: Re: throughtput measurements for fast ethernet
X-Priority: 3 (Normal)
References: <199705161709.KAA05644@conceptual.com> <12250.864080883@verdi.nethelp.no>
Content-Type: text/plain; charset=iso-8859-1
Sender: owner-hackers@FreeBSD.ORG
X-Loop: FreeBSD.org
Precedence: bulk

sthaug@nethelp.no wrote:

> > > I was hoping for something more specific. My _first_ tests with
> the 3com vx
> > > driver for the 395 pci card on FreeBSD 2.2.1 are not showing good
> results -
> > > highest throughput for UDP 45Mbps. I should stress that these were
> first tests.
> > >
> > > Has anyone any traffic measurements for any of the fast ethernet
> drivers?
> >
> > What kind of main memory bandwidth can your motherboards sustain?  A
>
> > good test is the stream benchmark from McCalpin.  Many P6
> motherboards
> > can sustain transfer rates which drive 100mb ethernet at full
> throttle,
> > but a lot of P5 motherboards cannot.  45Mbps is not unusual for
> these.
>
> I doubt motherboards are the issue here. I've measured 79 Mbps between
>
> two P133s (430FX chipset), using SMC (21140 based) cards. This was
> done
> more than a year ago, and the results are in the Netperf database. I
> got
> the same figures with ttcp.
>
> Steinar Haug, Nethelp consulting, sthaug@nethelp.no

 Well, it does matter.  These had pipelined burst SRAM, right?  Anything

with async SRAM will run around 45Mbits/s with 21140 cards.  The
same CPU with PB-SRAM agrees with what you report.

However, she mentioned that other interfaces weren't affected, so in
this
case it's not the motherboard, and since I haven't tested the throughput

for the 3com driver I can't say for sure that it *isn't* the driver.

Russell