From mboxrd@z Thu Jan 1 00:00:00 1970 From: Benjamin LaHaise Subject: Re: Poor gige performance with 2.4.20-pre* Date: Sun, 29 Sep 2002 20:45:10 -0400 Sender: netdev-bounce@oss.sgi.com Message-ID: <20020929204510.A26826@redhat.com> References: <200209282257.g8SMvta32527@vindaloo.ras.ucalgary.ca> <002f01c2675d$b642b640$f5f2010a@weixl> <200209290634.g8T6Y2o08439@vindaloo.ras.ucalgary.ca> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Cc: "Xiaoliang (David) Wei" , netdev@oss.sgi.com Return-path: To: Richard Gooch Content-Disposition: inline In-Reply-To: <200209290634.g8T6Y2o08439@vindaloo.ras.ucalgary.ca>; from rgooch@ras.ucalgary.ca on Sun, Sep 29, 2002 at 12:34:02AM -0600 Errors-to: netdev-bounce@oss.sgi.com List-Id: netdev.vger.kernel.org On Sun, Sep 29, 2002 at 12:34:02AM -0600, Richard Gooch wrote: > This is all on a LAN (of course; expecting good performance from a WAN > is pretty futile). I use a buffer size of 256 KiB. >>From my experience tuning on a 550MHz P3 Xeon, you're better off using a buffer size of 8-16KB that stays in the L1 cache. Of course, that was without actually doing anything useful with the data being transferred. Gige really does need a faster cpu in the ghz+ range. As for ns83820, it's a work in progress. Some of the recent bugfixes may have reduced performance, so it may need to be retuned. -ben -- GMS rules.