From owner-freebsd-hardware  Wed Oct  9 15:24:31 1996
Return-Path: owner-hardware
Received: (from root@localhost)
          by freefall.freebsd.org (8.7.5/8.7.3) id PAA12924
          for hardware-outgoing; Wed, 9 Oct 1996 15:24:31 -0700 (PDT)
Received: from saguaro.flyingfox.com (saguaro.flyingfox.com [204.188.109.253])
          by freefall.freebsd.org (8.7.5/8.7.3) with SMTP id PAA12919
          for <hardware@FreeBSD.ORG>; Wed, 9 Oct 1996 15:24:27 -0700 (PDT)
Received: (from jas@localhost) by saguaro.flyingfox.com (8.6.12/8.6.10) id PAA26723; Wed, 9 Oct 1996 15:22:18 -0700
Date: Wed, 9 Oct 1996 15:22:18 -0700
From: Jim Shankland <jas@flyingfox.COM>
Message-Id: <199610092222.PAA26723@saguaro.flyingfox.com>
To: John.McLaughlin@acucobol.ie, rgrimes@gndrsh.aac.dev.com
Subject: Re: Poor SMC Etherpower 10/100 transfer rates
Cc: hardware@FreeBSD.ORG
Sender: owner-hardware@FreeBSD.ORG
X-Loop: FreeBSD.org
Precedence: bulk

Rod Grimes <rgrimes@gndrsh.aac.dev.com> writes:

> I have come accross something similiar in the very early days of
> the SMC9332DST.  It turned out to be one of the cards was bad in
> 10MB/s mode.  Note that this bad card drug the whole network
> down to <500kb/s until I found it and eliminated it.
> 
> The lesson learned from this was that just swapping cards does
> not always tell you that the cards are okay, it maybe one bad
> card dragging the setup down :-(.
> 
> The defective card in my case was not listening to the wire
> before starting to transmit, this caused excess collisions on
> the network and everyone suffered.  Even very light traffic from
> this node would cause serious problems for all other nodes, a
> simple ``ping hostname'' would drive my network peformance down
> the tubes :-(

By the most amazing coincidence (seriously), I just returned
from a visit to a client, where I was trying to debug some
network problems.  They have about 6 hosts on an Ethernet, but 3
of them are responsible for virtually all the traffic:  a
FreeBSD box with the SMC9332DST in 10 Mbit mode, and 2 WindowsNT
machines with Intel Ethermumble cards.  Total traffic on the
Ethernet was running about 2 Megabytes per minute (i.e.,
ballpark 3% of the theoretical maximum capacity); the FreeBSD
box is reporting a 20%+ collision rate!  I will now definitely
try swapping out the SMC card.

Just another data point ....

Jim Shankland
Flying Fox Computer Systems, Inc.