Recommendation for 1000BASE-SX card?
tedm at toybox.placo.com
Sat Jun 3 16:39:23 PDT 2006
>From: Danial Thom [mailto:danial_thom at yahoo.com]
>Sent: Saturday, June 03, 2006 4:26 PM
>To: Ted Mittelstaedt; Heinrich Rebehn; questions at freebsd.org
>Subject: RE: Recommendation for 1000BASE-SX card?
>I think its often difficult to distinguish
>between what is crappy, because good code can
>make bad hardware look good and vice versa. All
>ethernet controllers were designed by idiots.
>My first success story (now I don't want to let
>on to who I really am so I'll be vague), was an
>ISA card by a major vendor that locked up
>regularly, and it had a hideous reputation as
>being a bad card. It was the only card of its
>kind, and I needed it badly. They gave me
>schematics and said that they had tried and tried
>but couldn't find anything wrong with the card.
>They had contracted out to some brainfarm to
>write a driver, and the thing was this beautiful
>self-contained scheduler (this is like MSDOS 3
>mind you) with documented source, the whole deal.
>Well I tore it apart, simplified the code, got
>rid of all the soft interrupt passes and cleaned
>up all the memory management code. Now the card
>worked like a charm, didn't lock up, ran better
>than their spec and Mega-Billon$- company
>couldn't believe that some 23yo kid wrote a
>driver that a company they paid 100K to couldn't
>get to work.
Musta been one of those Intel SatisFAXion cards. ;-)
>My point is that until someone writes a really
>good driver you never know if hardware is any
>good or not. Now some hardware is hopeless. I'm
>not sure that the broadcom controllers are that
>hopeless. But since the intel cards work well and
>are cheap, who's going to spend the time to pour
>over the broadcom driver and make it better?
>There's a ton of I/Os in there that can be
>streamlined. But who's gonna do it? Its sure not
>worth my time.
More information about the freebsd-questions