From owner-freebsd-questions Thu Apr 11 13:01:34 1996 Return-Path: owner-questions Received: (from root@localhost) by freefall.freebsd.org (8.7.3/8.7.3) id NAA06474 for questions-outgoing; Thu, 11 Apr 1996 13:01:34 -0700 (PDT) Received: from rocky.sri.MT.net (rocky.sri.MT.net [204.182.243.10]) by freefall.freebsd.org (8.7.3/8.7.3) with SMTP id NAA06461 for ; Thu, 11 Apr 1996 13:01:30 -0700 (PDT) Received: (from nate@localhost) by rocky.sri.MT.net (8.6.12/8.6.12) id OAA04945; Thu, 11 Apr 1996 14:01:06 -0600 Date: Thu, 11 Apr 1996 14:01:06 -0600 From: Nate Williams Message-Id: <199604112001.OAA04945@rocky.sri.MT.net> To: bugs@freebsd.netcom.com (Mark Hittinger) Cc: questions@freebsd.org Subject: re: Slow 3c590's In-Reply-To: <199604111839.NAA03641@freebsd.netcom.com> References: <199604111839.NAA03641@freebsd.netcom.com> Sender: owner-questions@freebsd.org X-Loop: FreeBSD.org Precedence: bulk > I'm not sure this tid bit applies to the 3c590. > > The other 3c5xx cards I have played with have a firmware setting labelled > "client mode" and "server mode". > > Naturally I selected "server mode". Many FreeBSD'ers have made the same > choice for obvious reasons! > > As it turns out, the "server mode" is a way that 3com artificially throttles > the number of interrupts you can get per unit time (somehow). This is to > allow a novell style server to have some cycles for other things. > > If you set the card to "client mode" then the 3c5xx cards will process > interrupts as quickly as they can, without a firmware inhibitor. > > So boot the dos setup tool that comes with your ethernet cards and make sure > that you've selected "client mode" on all the card (assuming such a setting > exists on the 3c590). > > Your performance should go from dismal to tolerable. :-) Be careful. By setting this and a high baud rate (which performs a similar task of limiting the # of interrupts) you can cause the card to go 'too fast' for FreeBSD, causing the # of mbufs to quickly be run out. This happened on my 3C589B (PC-CARD), so I set the speed to 38400 which caused it to throttle itself because it thought the machine needed to service alot of serial interrupts. Nate