Re: Network Cards

[Nick Holland <nick@holland-consulting.net>]

> Huh? xl suck compared to fxp.

I think it is worth noting that xl(4), fxp(4), ahc(4), etc. are
DRIVERS, not cards.

There are a lot of different cards that use the fxp(4) driver, several
that use the xl(4) driver, gobs that use the ahc(4) driver, and often
with radically different personalities.  You really got to talk about
the precise card and revision, not the drivers.

I'm partial to fxp(4) cards myself, I'd love to tell you what ones,
but unfortunately Intel doesn't seem to think it fit to put useful
model numbers on their cards.  Or are these even Intel-made cards? 
(some I have have an IBM sticker on 'em).  The old fxp(4)s I have are
nice, the newer ones have an on-board boot ROM which is annoying as
heck -- it slows the boot process by several seconds (the delay can be
reduced (usually -- saw one that wouldn't let you reduce it...20
second added boot time), but never eliminated, at least not without
unsoldering the ROM chip from the board), I'd avoid putting them in
workstations, but they work fine in "rarely booted" systems (i.e.,
OpenBSD servers/firewalls/etc.).  Of course, if I were doing PXE
booting, these cards would go from detestable beasts to treasures very
quickly in my mind.  Some people have reported serious problems with
some fxp variant, which apparently I have never come across.

I've had clients with massive problems with rl(4) cards (in
non-OpenBSD networks) which I solved by replacing the all the rl(4)
cards with ... rl(4) cards.  The $25 card was a nightmare,
malfunctioning in most of the workstations it was in, the $8 card
worked great for several years after I installed them.  Go figure.


Me?  I buy lots of network cards surplus, for $2 to $5US ea.  Test 'em
out, get to know 'em, use 'em where appropriate.  For most people,
most cards will work fine most of the time.  If you really have an app
where you have to pick the "best" card, you can afford to buy three or
four different cards and try 'em out in YOUR application.

Try 'em.  If it works, great.  If it doesn't, try another.  If that
one works, you may wish to stock up, who knows what they will be
sticking the same label on tomorrow...  *sigh*

Nick.
-- 
http://www.holland-consulting.net