>4.  The 2 cards seem to have different chips on them, but I don't remember
>what they are, given that the cards aren't in my posession.

Uh, oh. If 3Com have redone the ASIC, anything could be going wrong.

First thing is to read the 16-bit words at offsets
EEPROM_{SOFTINFO,SOFTINFO2,CAP} from the eeprom (see how elinkxl.c
reads the oemaddr), figure out which bits are set and what they mean,
or at least if the values on the two cards differ.

(I think there's 32 bits of capabilities altogether, but I'm not sure
where the other 16 bits are. Perhaps EEPROM_CAP+1? My refs are packed
but i can check over the weekend if noone else does.)


>2.  Is it possible this card, which is brand new, is one of the new
>3c905b-M cards which is being mis-identified with a 3c905b-TX?

Possibly. Historically 3com have been good about bumping the PCI product-ID
if they make software visible changes, so its unlikely. But possible.

>5.  Do the DOS/Windows configuration settings for performance versus CPU
>load in the card's diagnostic program affect the NetBSD's driver behavior
>toward this card?

Um. If memory serves, its up to each OS's driver to read these values
when it attaches the card, and set interrupt hysteresis levels accordingly.