Problems with LinkSys 10/100 cards

Robert G. Brown rgb@phy.duke.edu
Fri Feb 5 14:45:12 1999


On Fri, 5 Feb 1999, Marko Kiiskila wrote:

> 
> > > Netperf seems to be much more reliable but I have seen dependency on
> > > what I run and when, i.e. netperf first or netpipe first, etc...
> > > 
> > > Anyway I managed to find a copy of the datasheet for the PNIC part, so
> > > maybe I can make more headway.
> > > 
> > > I have seen transfer rates as high as 67Mbit/s, when things are
> > > working, so I guess the cards are working properly ??
> > With netperf I get 95-96 Mbit/sec with both UDP and TCP.
> > With netpipe it seems to top about at about 91 Mbit/s
> 
> And what was the packet size in these tests? Not small packets, I assume?
> I'd be very interested in seeing the measurements done with 64byte, 128 byte,
> 256, 512, 1024 and 1476 packet sizes. 

Yes, note well that measured transfer rates depend strongly on CPU speed
and packet size for packets less than around 1K in size.  There is a
graph, and set of instructions for measuring performance with netperf,
on www.phy.duke.edu/brahma (the graph is even for a tulip card, although
not a linksys).  Measured throughput will also depend strongly on the
CPU load on the receiver (and other things, but these three are the
primary determinants).

   rgb

Robert G. Brown	                       http://www.phy.duke.edu/~rgb/
Duke University Dept. of Physics, Box 90305
Durham, N.C. 27708-0305
Phone: 1-919-660-2567  Fax: 919-660-2525     email:rgb@phy.duke.edu