From owner-freebsd-hackers Mon May 3 16:22:11 1999 Delivered-To: freebsd-hackers@freebsd.org Received: from implode.root.com (root.com [209.102.106.178]) by hub.freebsd.org (Postfix) with ESMTP id 173EE154AC for ; Mon, 3 May 1999 16:22:08 -0700 (PDT) (envelope-from root@implode.root.com) Received: from implode.root.com (localhost [127.0.0.1]) by implode.root.com (8.8.8/8.8.5) with ESMTP id QAA04559; Mon, 3 May 1999 16:18:40 -0700 (PDT) Message-Id: <199905032318.QAA04559@implode.root.com> To: Wes Peters Cc: sthaug@nethelp.no, hackers@FreeBSD.ORG Subject: Re: Decent network cards for 100Mbit? In-reply-to: Your message of "Mon, 03 May 1999 16:56:55 MDT." <372E29B7.2102CC39@softweyr.com> From: David Greenman Reply-To: dg@root.com Date: Mon, 03 May 1999 16:18:40 -0700 Sender: owner-freebsd-hackers@FreeBSD.ORG Precedence: bulk X-Loop: FreeBSD.ORG >sthaug@nethelp.no wrote: >> >> > My only complaint about the EEPro 100B cards is paying Intel $65 for >> > a card that has a single $4 chip on it. Bleh. Plus, the performance >> > I've seen hasn't been all that stellar, but I may be doing something >> > wrong. I haven't really tried tuning the system much yet, just doing >> > some simple throughput tests using ftp and tcpblast. >> >> Performance from these cards should be very good. I was able to receive a >> full 100 Mbps with one of these cards in a P-133 running 3.0-970124-SNAP, >> ie. more than two years ago. >> >> To *send* a full 100 Mbps you need slightly more CPU, say a P-166. > >I should probably point out I'm doing network throughput torture >tests with 64-byte packets. ;^) Any reasonably good Fast Ethernet >card on a respectable PCI machine will do 100 Mbps using 1500 byte >packets. Using FreeBSD-3.1 (somewhere between RELEASE and STABLE) >on a PII/233 (sender) and PPro 200 (receiver), I get the following >figures: > > Driver Card Throughput > ------ ------ ------------------------------------- > xl 3c905 30.4 Mbps > xl 3c905B 32.0 Mbps > doesn't work unless there is a 3c905 > in the system as well. > fxp EEPro 100B 32.7 Mbps > pn LNE100TX 39.6 Mbps > >I haven't attempted to measure interrupt overhead or anything like >that since the only important measure for MY needs are "how fast >can this thing stuff bits onto the wire?" I think the pn device is >approaching the saturation point; the others do not come close. > >I'll also be using these for testing multicast throughput later on, >with a lot of multicast groups. Since the PNIC chip supports "perfect >multicast filtering," I.e. it has hardware filters for all ethernet >multicast addresses, I am hoping these cards will perform well in this >environment as well. > >Your mileage may vary. ;^) You benchmark is broken. The pro/100, for one, can easily do 100Mbps continuously. I should know since not only did I write the driver, but I use it exclusively here and on wcarchive (and in the latter, I have graphs and download stats to prove it). -DG David Greenman Co-founder/Principal Architect, The FreeBSD Project - http://www.freebsd.org Creator of high-performance Internet servers - http://www.terasolutions.com To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-hackers" in the body of the message