i would imagine the fsb restriction is what the cpu passed in the testing phase before being badged whatever designator at the end. if the new barton chip isn't upto spec stability at 400, but is at 333, it'll prob get the 333 badging at whatever speeds. now that doesn't mean that it can't run at 400 fsb, it just means that AMD doesn't guarantee 400fsb perf.
same thing for the overclockable chips, their safe speeds are what they badge the chips, but some may have more potential for OC than others, like in the case of XP1700+ XP2100+ and in P4, the upcoming 2.4c 800mhz pre production model got cranked upto 3.3ghz w/ stock cooling.. amazing
but more importantly, (something that i picked up reading the intel 800fsb article) your fsb needs to be in sync w/ the ddr memory to get best performance. if your ram can do ddr 333, then you are better left w/ a 166 fsb. (166 x 2 = 333 ddr)
my answer (opinion) to the original question would be if all other components were the same, whatever fsb matchs up with the ddr ram x multiplier will yield the highest perf at the same final clock speeds.
Facebook
Twitter