Skip site navigation (1)Skip section navigation (2)
Date:      Fri, 18 Oct 2002 09:07:40 -0700 (PDT)
From:      John Polstra <jdp@polstra.com>
To:        stable@freebsd.org
Cc:        jamie@jamiesdomain.org.uk
Subject:   Re: Compaq GIGA NIC
Message-ID:  <200210181607.g9IG7eGK028163@vashon.polstra.com>
In-Reply-To: <001501c275eb$a9328560$3264a8c0@BONG>
References:  <FE045D4D9F7AED4CBFF1B3B813C8533701022CB1@mail.sandvine.com> <001501c275eb$a9328560$3264a8c0@BONG>

next in thread | previous in thread | raw e-mail | index | archive | help
In article <001501c275eb$a9328560$3264a8c0@BONG>,
Jamie Heckford <jamie@jamiesdomain.org.uk> wrote:
> > > From: Jamie Heckford [mailto:jamie@jamiesdomain.org.uk]
> > >
> > > Urgh - The broadcom chips perform terribly!
> >
> > I'm curious why you say that. I have a large number of both
>
> Last time we tried this driver (around 4.6-RELEASE) we had massive
> problems with the chipset occasionally incorrectly checking TCP
> checksums... which caused the interface to lock up.
>
> I think someone posted a workaround to disable the hardware checksum
> checking feature - but we only had these cards onboard in our
> production servers.
>
> Maybe its corrected now... who knows?? By the sounds of it you
> aren't having any issues :-)

So you ran into a driver bug, and you describe it by saying "The
broadcom chips perform terribly!" ???  No wonder there's so much
misinformation on the net.  Sheesh, try to remember that this stuff
gets archived!  People read it and believe it, and nobody benefits
when it's just not true.

I've been working with both the various models of Broadcom and Intel
gigabit NICs for close to 2 years in a setup where it's possible to
measure their performance accurately under controlled conditions.
The Broadcom chips perform on a par with the Intel chips under all
circumstances.  With certain kinds of traffic, the Broadcom chips
perform measurably better.

If you are running into trouble with the Broadcom chips, it's because
of _bugs_ in the driver (or because of the lack of the necessary
work-arounds for chip bugs).  Those bugs exist because, just like
Intel, Broadcom won't release the technical details of the chips
without an NDA.

Don't confuse driver problems with chip problems.

John
-- 
  John Polstra
  John D. Polstra & Co., Inc.                        Seattle, Washington USA
  "Disappointment is a good sign of basic intelligence."  -- Chögyam Trungpa


To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-stable" in the body of the message




Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?200210181607.g9IG7eGK028163>