From owner-freebsd-hackers Sun Mar 14 3:38:35 1999 Delivered-To: freebsd-hackers@freebsd.org Received: from mw5.texas.net (mw5.texas.net [206.127.30.15]) by hub.freebsd.org (Postfix) with ESMTP id 4495B14BE4 for ; Sun, 14 Mar 1999 03:38:32 -0800 (PST) (envelope-from rsnow@lgc.com) Received: from basil.dympna.COM (mnet06-30.sat.texas.net [209.99.48.198]) by mw5.texas.net (2.4/2.4) with ESMTP id FAA01122; Sun, 14 Mar 1999 05:38:12 -0600 (CST) Received: from solo (solo [134.132.228.3]) by basil.dympna.COM (8.9.1/8.8.7) with SMTP id FAA30428; Sun, 14 Mar 1999 05:38:10 -0600 (CST) (envelope-from rsnow@lgc.com) Message-ID: <001b01be6e10$08061120$03e48486@dympna.com> From: "Rob Snow" To: "Amancio Hasty" , "Cory Kempf" Cc: "Bill Paul" , Subject: Re: Gigabit ethernet -- what am I doing wrong? Date: Sun, 14 Mar 1999 05:44:28 -0600 MIME-Version: 1.0 Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: 7bit X-Priority: 3 X-MSMail-Priority: Normal X-Mailer: Microsoft Outlook Express 5.00.0810.800 X-MimeOLE: Produced By Microsoft MimeOLE V5.00.0810.800 Sender: owner-freebsd-hackers@FreeBSD.ORG Precedence: bulk X-Loop: FreeBSD.ORG I'm looking at this wondering a couple of things. How many mem copies take place in the IP stack before we're ready to transmit a frame? My question is based around whether it's the NIC's or the IP stacks and PCI holding us back. What would PCI-64@66 do for us with current stacks? -Rob ----- Original Message ----- From: Cory Kempf To: Amancio Hasty Cc: Bill Paul ; Sent: Saturday, March 13, 1999 11:22 PM Subject: Re: Gigabit ethernet -- what am I doing wrong? >Amancio Hasty writes: > >> > 200 Mb/s = 25 MB/s, which seems a little low, but is within the realm of >> > what I would expect. >> >> I think the system should be able to support at least 70MB/s at least I >> do over here >> with a bt848 video capture board capturing 640x480x4 at 30 frames per second >> and then displaying the frames on video display card 8) > >A video capture board is generally moving bulk data without any protocol >in the way. It is idealy suited for getting maximal bandwidth from PCI, >as you can essentially set up the transfer, then just let it run. > >With an ethernet driver, though, there is often additional host<->card >traffic, such as telling the card "here is some data", the card responding >"Ok, I am done with it", etc. Additionally, the protocol doesn't lend >itself to exclusively bulk data transfers. ARP and other overhead will >eat up a lot of that theoretical bandwidth. > >So, if you are seeing 70 MB/s with video, where you can probably get by >with a single ACK between frames -- i.e. 2 bus operations / block -- it >wouldn't surprise me to see an ethernet card, with a protocol that might >require five bus operations / block getting less. > >Remember too, that trip through the protocol stack was only a little >over 50 MB/s... > >+C > >-- >Thinking of purchasing RAM from the Chip Merchant? >Please read this first: > >Cory Kempf Macintosh / Unix Consulting & Software Development >ckempf@enigami.com > > >To Unsubscribe: send mail to majordomo@FreeBSD.org >with "unsubscribe freebsd-hackers" in the body of the message > > To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-hackers" in the body of the message