Testing the GNIC-II (Hamachi) gigabit NIC

[Jes Sorensen]

>>>>> "Jan" == Jan Rychter <jwr@rgstudio.com.pl> writes:

>>>>> "Jes" == Jes Sorensen <Jes.Sorensen@cern.ch>:
Jes> Right, for me with the Alteon card, the performance increases from
Jes> like 28MB/sec to around 90MB/sec in a P2 400MHz 440BX (PCI
Jes> 32/33)box when going from 1500 byte frames to 9K ones.

Jes> However, I would be interested to hear what kind of performance you
Jes> get on 1500 byte frames.

Jan> 31MB/s and that's doing ftp transfers of a large file (~80MB) to
Jan> /dev/null. The transfers were from an Alpha 533 (128MB RAM) to a
Jan> Pentium-II 266.

Jan> 31MB/s TCP is impressive in itself, but I think both the card and
Jan> the machines are capable of more.

I tend to disagree on that, we are talking about a media which is 10
times as fast as Fast Ethernet and we only see a 2.5-3 times performance
increase. IMHO that is very disappointing, in particular when we know
that the same machines can go much much faster with other network media
or even the same media with larger frames.

As an example I get 75MB/sec in a Pentium II/300 when using 9K frames
and only ~28MB/sec when using 1500 byte frames. With a better PC I get
as much as 90MB/sec and I hope/expect it will go even further when I get
my hands on a box with a 64-bit PCI bus.

I think I may be able to get a bit more out of the card with small
frames as well, but I think the people designing the GigE standard major
a major mistake when they decided to limit it to the pre-historic 1500
byte frames.

Jes
 | To unsubscribe, send mail to Majordomo@cesdis.gsfc.nasa.gov, and within the
 |  body of the mail, include only the text:
 |   unsubscribe this-list-name youraddress@wherever.org
 | You will be unsubscribed as speedily as possible.