I got my SMC EtherPower II 10/100 (with WakeOnLan :-} ) and put it in my
PM 7600. This is the same machine with the D-link (rtk) and Asante
(was de, now tlp) cards that I've been reporting on.

The epic driver recognizes the SMC card, and selects 100BaseFX right
out of the box.  Stress testing it yields better performance than either
the D-link or Asante cards (regardless of whether using de or tlp)

Reported below is some of my impressions.  I did the performance testing
by doing ftp transfers of real disk data, and I know that my SCSI drive
won't put out 10 MB/s, so the performance thing is a tad bogus, but the fact
that I get different numbers from the different cards indicates that
there are efficiency differences of some sort.

It looks to me like:

	o the D-link is reliable, but slow.  I get no whines in the
	logs, and no hiccups on data xfer, but data rates are only
	about 1.5 MBytes/sec

	o Asante is a little flaky.  It doesn't seem to matter much whether
	I use tlp or de.  Neither one will go 100 BaseT without manual
	intervention (ifconfig ...media...)
	At 10BaseT, it's pretty solid, but at
	100BaseT, it seems to "stall" frequently, causing poor
	performance.  I also get some log messages the first time I
	stress the card about buffer underrun - new buffer size - store
	and forward mode.  I presume this is as it should be?
	Resultant performance is about 1.2 to 1.5 Mb/s

	o SMC Etherpower I board is found by the epic driver,
	selects 100BaseT without prodding, and delivers a steady
	2.2 MB/s (limit of disk speed, probably) what appears to
	be all day long.

Next step would be to get another machine with  known-to-be-good-and-fast
100BaseT that I can use to run some *real* tests....

-dgl-