Well I must say I noticed a bit of a difference when I built a couple of Athlon 64 boxes for clients. Since only XP Pro (32bit) was installed I can only give you a subjective opinion on the speed. However I think that there was a noticable difference with the Athlon 64 box compared to a both a P4 and Athlon XP box of similar speeds. All things being equal, I think that perhaps there is also some differences in both the CPU and the Motherboard chipset that seems to help out here, even on a 32bit OS.
The motherboard was the
Asus K8V based on the VIA KT800 chipset. I beleive the effeciency of the CPU/Chipset and motherboard design help to increase the performance even on a 32bit OS

Of course I didn't get to play around as long as I wanted to but 3D Mark also gave higher scores then comparable boxes with other CPU's.
Also note that the Athlon 64/FX share the same heatsink/fan combo as the Opteron does. Nice way of keeping cost's down so you don't have to come out with separate cooling devices for each flavor of CPU
