Re: [uCsimm] eth circuits: are there 100mbit circuits too?

From: Geoffrey Wossum (gpw0341@omega.uta.edu)
Date: Thu Oct 05 2000 - 21:59:39 EDT


> Well, my uCsimm can't even push maxing out the 10bt interface so there
> is really no point in adding anything faster. Choosing a 100bt chip
> over a 10bt would only add extra cost not extra speed.

Yes, that sounds right. If you think about, the 'ez328 runs at 2.77
bogoMIPS (if you'r not using the LCD controller). And if memory serves
correct, it's connected to the CS8900A by a 16bit data bus. So even if it
wrote a 16 bit word to the Ethernet controller every instruction, it could
only push out 43.2 Mbit.

And we all know you can't write a word to the NIU every cycle. You have
to copy it out of a buffer, increment the buffer pointers, etc. You could
only hit 10 Mbit if you could a push a word to the NUI every fourth cycle,
which I think everyone would agree is totally unrealistic. And then you'd
be doing nothing but handling network traffic, you wouldn't be running
processes to generate network traffic.

Anyway, a lot of PC-class machines (386's, 486's and low-end Pentiums) can
be flooded by even a 10Base connection.

---
Geoffrey Wossum
gwossum@acm.org
Project AKO - http://rover1.uta.edu/~ako
Internet Imperialists - http://inetimperial.sourceforge.net
This message resent by the ucsimm@uclinux.com list server http://www.uClinux.com/



This archive was generated by hypermail 2b30 : Sun Apr 07 2002 - 00:01:38 EST