[tulip] Re: McQullian hacked 0.91g-ppc tulip.c driver for AN985 chipset --WAS: Linksys LNE100TX Version 4.1

Donald Becker becker@scyld.com
Mon, 7 Aug 2000 00:23:25 -0400 (EDT)


On Sun, 6 Aug 2000, Bryan J. Smith wrote:

> > No!  It's more than just an identification issue.  The Comet/Centaur chip
> > isn't exactly like other tulip-like chips.  (Not that any of them are like
> > the others.)
> 
> Hmmm, better?  worse?  How the CPU utilization (i.e. would you use
> this in a server)?

We are using them in our test cluster.
They should be nominally about the same as a real Tulip:
   - Recieved frames must be put into aligned buffers, which mis-aligns the
     IP header. 
   - There is no UDP/TCP/IP checksumming on the chip

The differences are in the media selection and EEPROM layout.  An advantage
of a single-chip implementation is that is only one type of transceiver
connection to test against, unlike the 21140 and 21143 which had hundreds of
variations.

Donald Becker				becker@scyld.com
Scyld Computing Corporation		http://www.scyld.com
410 Severn Ave. Suite 210		Beowulf-II Cluster Distribution
Annapolis MD 21403