The 333MHz FSB is actually only a 166MHz FSB in actual clock speed; but it is double pumped (i.e. DDR) so it gets two transfers per clock signal: 166 x 2 = 332. But 333Mhz looks better, so they rounded it up.
AMD's current processors use a 266MHz FSB (which is really only a 133MHz clocked bus, with dual tranfers). Earlier processors were 200MHz FSB (again, really only 100MHz clock).
So, the 166MHz FSB (333MHz effective rate) seems to be the next logical step for increasing the clock speeds of things. No, they don't officially have it yet, but that is where they are going and that is what the SiS735 chipset ALREADY supports...again, just not totally official. (Although this news bit pretty much makes it official...) DDR memory is also moving to that same clock rate. There is a spec out there in Standards Land for DDR333 SDRAMs (may not be finalized yet, though). (So, is 333MHz the next 'technically' logical step, or did AMD and memory standards bodies agree to make that the next step? If they did, I would assume that there was technological reason behind it.)
I won't get into the P4's bus speed because I just had second thoughts about what I was going to say. I believe that they use the same kind of trick with their bus speeds: the specified speed is usually 'effective speed' (the multiplied result) instead of the actual clock speed. (Rambus memory works at a 200MHz bus that is quad pumped and therefore gets rated as 800MHz RDRAM. At least the 800 stuff does. The slower versions - 700 and 600 - use a slower clocked bus that is still quad pumped.) But I wouldn't want to post something inaccurate and get flamed to death.
Justin