From owner-freebsd-hardware  Sun Aug 10 21:04:49 1997
Return-Path: <owner-freebsd-hardware>
Received: (from root@localhost)
          by hub.freebsd.org (8.8.5/8.8.5) id VAA07519
          for hardware-outgoing; Sun, 10 Aug 1997 21:04:49 -0700 (PDT)
Received: from databus.databus.com (databus.databus.com [198.186.154.34])
          by hub.freebsd.org (8.8.5/8.8.5) with SMTP id VAA07487;
          Sun, 10 Aug 1997 21:04:29 -0700 (PDT)
From: Barney Wolff <barney@databus.com>
To: freebsd-hackers@FreeBSD.ORG, freebsd-hardware@FreeBSD.ORG
Date: Sun, 10 Aug 1997 23:52 EDT
Subject: Re: question about "ed" driver performance on ASUS SP3G & 486DX4/100
Content-Type: text/plain
Message-ID: <33ee8f490.55c9@databus.databus.com>
Sender: owner-freebsd-hardware@FreeBSD.ORG
X-Loop: FreeBSD.org
Precedence: bulk

As a data point, I just copied 5MB from a P6/200 running fbsd 2.1.5 & a
Pro 100B (at 10 MHz) to a 486/66 running Unixware 2.0.3 with a WD8013.
Via NFS, 651 KB/sec.

I have in fact seen slow transfers to a 486 with WD8013 from an SGI
workstation.  I wonder if it has more to do with the ethernet than
with the 8013.  But on the same lan, another 486 with a 3com 3c509
was able to keep up.  It may have something to do with how collisions
get handled, as I vaguely recall (was a couple of years ago) that
running multiple transfers simultaneously caused trouble with the 8013
but not with the 3c509.

WARNING:  In all the tests I've run, the 486 was running Unixware, not
fbsd.

<barney@databus.com>


> Date: Sun, 10 Aug 1997 19:32:27 -0700
> From: Mike Haertel <mike@ducky.net>
> 
> I have two boxes, one based on the ASUS SP3G with a 486 DX4/100
> processor, and another based on the ASUS P6NP5 with a 150 MHz PPro.
> The 486 box has a 16-bit WD8013 based ethernet board, and the PPro
> has an EtherExpress Pro/100.
> 
> I attempted to do an NFS install of FreeBSD 2.2.2 on the 486 box
> using the PPro box as the server.  At appeared to detect the
> ethernet board Ok, but it got hung when actually trying to copy
> files.  After a considerable pain I concluded that it was dropping
> the trailing packets (fragments), and the @#%@! UDP and/or
> NFS protocol on the server was responding by attempting to
> retransmit the entire packet again, and thus causing the trailing
> packets to be lost again.  It seems that the PPro pumps the bits
> out on the wire so fast that the 486 had no time to catch its breath.
> Setting the maximum NFS read size to 2K or smaller allowed it
> to work.  But slowly.
> 
> The same wd8013 ethernet card worked fine for a network install
> to a Pentium/90 based Intel Xpress box.  I really have trouble
> believing the 486/100 is so much slower than the Pentium/90
> it can't keep up.
> 
> So: Is there anything special I should know about wd8013 cards
> and ASUS SP3G's and/or 486/100's?  Or am I just plain out of luck?
> In the latter case could anybody recommend a faster ISA ethernet
> card that's widely supported by the free OS's?
> 
> Thanks,
> 
> 	Mike
>