udelay(500) in amd8111 and nforce2

[mds at paradyne.com (Mark Studebaker)]

This is definitely no good.
The preferred implementation is #if'ed out in nforce2.
Does somebody have these chips to test the fix or
should we just change it w/o testing first?