RE: [ethmac] MAC Rx architecture

["Kevin Kay" <kevin.kay@espipd.com>]

At a high-level, I think what you are saying is true.  I've not done a GMII
interface as all of my GigE experience to date has been fiber.  However, I
think the GMII interface is 8-bits wide (as opposed to 4-bits wide for MII),
and a GMII runs synchronous to a 125MHz clock (as opposed to 2.5MHz or 25MHz
for MII).  Seems like there would be a lot of issues to work out in scaling
multiple 10/100 MACs to act a Gig MAC.  CRC checking is the first thing that
comes to mind, MAC address comparison (checking for broadcast MAC addresses,
etc.) is another issue that would seem difficult to do without substantial
interaction between the MAC "sections."

I don't know your application, but I would think that you have to be very
close (if not over) a cost/performance crossover point such that you could
just go buy a Gig MAC/PHY and hardwire it to run like you want (granted, a
lot of the MACs will probably have a PCI interface on the host side, which
may or may not add un-needed complexity to your architecture).  Just some
thoughts from someone who knows zero about your specific application.

Regards,
Kevin

-----Original Message-----
From: owner-ethmac@opencores.org [mailto:owner-ethmac@opencores.org]On
Behalf Of Colin
Sent: Thursday, August 01, 2002 9:01 AM
To: ethmac@opencores.org
Subject: [ethmac] MAC Rx architecture



Having got some great feedback on this forum already, I'd like to ask
another question - one which I've been pondering for a couple of days
now:

If the functionality of the MAC Receiver for GigMAC is basically the same
as for Fast Ethernet (full-duplex only), what is the difference in the
resulting hardware architecture? Is it simply that for GigMAC the
hardware must operate faster to handle the faster arrival of data? Or
would a completely different approach be required?

I've thought about using an arbiter with multiple instantiations of Fast
Ethernet MAC Rx units to acheive Gigabit throughput as an option. At
the moment though I'm trying to go with a single rx unit operating
extremely quickly.

Any thoughts?

Cheers,
Colin

--
To unsubscribe from ethmac mailing list please visit
http://www.opencores.org/mailinglists.shtml

--
To unsubscribe from ethmac mailing list please visit http://www.opencores.org/mailinglists.shtml