Skip site navigation (1)Skip section navigation (2)
Date:      Sun, 20 Oct 2002 19:08:38 -0400
From:      Don Bowman <don@sandvine.com>
To:        'Petri Helenius ' <pete@he.iki.fi>
Cc:        "'freebsd-net@FreeBSD.ORG '" <freebsd-net@FreeBSD.ORG>
Subject:   RE: ENOBUFS
Message-ID:  <FE045D4D9F7AED4CBFF1B3B813C8533701022CE6@mail.sandvine.com>

next in thread | raw e-mail | index | archive | help

From: Petri Helenius
>>Well, I'm definitely finding that I have more CPU free when using
>>the broadcom BCM570X NIC (bge) than the Intel 8254X NIC (em).

>What kind of difference are we talking about here? One third or less?
 ...

The idle time in a bridging application remains more or less
constant for the bge, regardless of load. This stayed at around
9% of my CPU. For the em, the idle time decreased as I increased
the load, to end up using about double the CPU for the same load.
I will post results when I have them done.

>I think I=B4ll put a 3com card onto the PCI-X slot on the same chassis
>and do some comparisons on the same exact hardware, traffic and kernel
>configuration.
>
>However, if I understand correctly, PCI-X would allow optimizations
>that are not present on the em driver?

I'm not sure what these optimisations would be other than clock
rate which the driver doesn't care about.

Watch that your GE cards use a 64-bit bus, and stay at at least
66MHz. For PCI-X, a single device can run @ 133, 2 @ 100, more than
2 @ 66MHz. In the supermicro servers I have, one of the 2 slots
is better than the other since its only used by the expansion.

--don (don@sandvine.com www.sandvine.com)

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




Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?FE045D4D9F7AED4CBFF1B3B813C8533701022CE6>