Date: Mon, 27 Feb 2006 17:19:50 -0600 From: Kevin Day <toasty@dragondata.com> To: Andrew Seguin <asegu_fbsdnet@borgtech.ca> Cc: freebsd-net@freebsd.org Subject: Re: Network card selection / recommendations ? Message-ID: <FF3DC3AF-64AF-426F-8391-24D99EFD18FB@dragondata.com> In-Reply-To: <44037A94.3030903@borgtech.ca> References: <44037A94.3030903@borgtech.ca>
next in thread | previous in thread | raw e-mail | index | archive | help
On Feb 27, 2006, at 4:17 PM, Andrew Seguin wrote: > First off, several gigabit network cards advertise various kinds of > network cable diagnostics (Intel "Advanced Cable Diagnostics", D- > Link "Cable Diagnostic",for example). We have in one building here > old wiring that wasn't professionaly installed and in most places > only 10mbps works although the wires are certified cat5. We don't > have the money for a good network tester with TDR to find out what > kind of problem exists (although a simple pin-out tester says there > are no crossed wires). So question number is really this: does > anybody have any experience with this kind of functionality on a > network card? is it worth anything? > Are you trying to run 100 or 1000mbps? If the latter, keep in mind two things. 1) 1000BaseTX requires all 4 pairs(8 conductors), not the 2 pairs that 10/100 requires. It's possible the remaining 2 pairs aren't wired correctly for gigabit speeds. 2) Technically GigE requires Cat5e or Cat6, but lots of people get away with Cat 5. 3) In a perfect world, GigE over copper can only go 100 meters. Less if you're not using the right cables, not using good connectors, etc. What are your lengths like? The built in cable tester thing is basically a pass/fail that isn't too accurate. Even if it did though, what information are you looking for? If it's not working... It's not working. :) There's either a cable/connector problem, or something wrong at the switch/NIC. > Second... is there any advantage to dual port cards in term of > performance? I always imagined there could be a solution which > could be a big performance boost if a packet came in one port, the > headers only went over the PCI bus for processing by the OS, and > then the firewall/gateway for example either tells the NIC to > delete the packet or forward it off the second port (all without > copying the entire packet into system memory). Is that just my > imagination of a perfect world or...? Nope. Every dual/quad/etc port I'm aware of is treated like two totally independent cards. Even the high performance Intel boards can't share packet data between ports. Would be nice though. :)
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?FF3DC3AF-64AF-426F-8391-24D99EFD18FB>