DDR SDRAM is nonetheless a very important technology.  it DOES offer a benefit (8% above KT133A is rather small), but not much, because as Anand stated, the Athlon is designed around SDRAM, so slow RAM is part of the game.
IF however, AMD releases the thoroughbred designed for DDR SDRAM, I wouldn't be surprised if it is better.. 
Recap, FSB is very important for the Athlon.  With Linpack, it SHOULD be RAM througput being the biggest problem..  thus, it leads me to conclude that, yes increased FSB is part of the solution, BECAUSE the CPU is doing more then just that one program (how would you then test it in a pure environment?  boot the computer to an OS you trust, in this case, you'd have to boot it to a testing OS, made only for this application, which is running tests, and nothing else.
RAM throughput is important when it is the limiting factor.  in this case, I don't think it is, because I think the FSB isn't fast enough to keep up with whatever else is going on BESIDES the RAM througput.
P4 has such a fast FSB, that it can do that kind of thing.
An Athlon with equivalent fsb speed (200mhz, 400mhz effective), as well as RAM speed, would undoubtably feature at, or near P4 speeds in Linpack.  with slower DDR SDRAM (say PC2100), the performance would definetly suffer in this situation, but with DDR SDRAM capable of 3.2 gig's a second, it would definaetly be on par with the P4, when paired with a 200mhz FSB (400mhz effective, for those who need to be clear).
this is the EXACT reason why I think AMD should ramp up FSB speeds all the way.  they most likely will anyway with the Thoroughbred version of the Athlon.
however, I could be wrong..