From owner-freebsd-questions Thu Apr 25 09:56:28 1996 Return-Path: owner-questions Received: (from root@localhost) by freefall.freebsd.org (8.7.3/8.7.3) id JAA21624 for questions-outgoing; Thu, 25 Apr 1996 09:56:28 -0700 (PDT) Received: from halloran-eldar.lcs.mit.edu (halloran-eldar.lcs.mit.edu [18.26.0.159]) by freefall.freebsd.org (8.7.3/8.7.3) with SMTP id JAA21619 for ; Thu, 25 Apr 1996 09:56:26 -0700 (PDT) Received: by halloran-eldar.lcs.mit.edu; (5.65/1.1.8.2/19Aug95-0530PM) id AA26011; Thu, 25 Apr 1996 12:56:15 -0400 Date: Thu, 25 Apr 1996 12:56:15 -0400 From: Garrett Wollman Message-Id: <9604251656.AA26011@halloran-eldar.lcs.mit.edu> To: Michael Smith Cc: freebsd-questions@freefall.freebsd.org Subject: Re: What network cards work well? In-Reply-To: <199604250835.SAA24453@genesis.atrad.adelaide.edu.au> References: <317E74AE@smtp> <199604250835.SAA24453@genesis.atrad.adelaide.edu.au> Sender: owner-questions@FreeBSD.ORG X-Loop: FreeBSD.org Precedence: bulk < said: > Robert Clark stands accused of saying: >> PCI NICS: >> What PCI bus NIC(s) has/have the most fully feature driver support >> under FreeBSD 2.1? > Probably the DEC-based cards. >> If I run a DEC based Fast Ethernet card, will it run well in a standard >> 10Mb network? Do the DEC based cards suffer when compared to a standard >> 10Mb card? > No. The DEC chipset is pretty good/ Some more information on Ethernet chipsets based on our experiments here... - The DC21140 is OK, but it has a couple of problems: 1) It makes you do unaligned accesses on received data. 2) The bus overhead is fairly large, and is greater on Orion chipsets (P6) than on Tritons. This means that my P6-150 actually transmits /slower/ than my P5-133 does. - The i82557 is OK, but it has a strange performance anomaly: The Intel chip (unlike the DEC one) is capable of running at full Fast Ethernet speeds for packet sizes greater than about 450 bytes. When the packet size is less than this, however, it straight-line falls off the curve. We have not yet determined whether this is actually the fault of the 82557 or not, but in any case it is disturbing. What this means as a practical matter is that if your packets tend to be small (less than 250 bytes), the DEC chip can transmit them faster. If your packets tend to be large, the Intel chip can transmit them faster. On a 10-Mbit network, these numbers are irrelevant, since they are far above the curve. I have not seen any AMD 100Mbit parts (if there even is such a thing). -GAWollman -- Garrett A. Wollman | Shashish is simple, it's discreet, it's brief. ... wollman@lcs.mit.edu | Shashish is the bonding of hearts in spite of distance. Opinions not those of| It is a bond more powerful than absence. We like people MIT, LCS, ANA, or NSA| who like Shashish. - Claude McKenzie + Florent Vollant