Well, I have found your article interesting and realistic. I think, though, that it would be fair to compare single channel solutions against single channel solutions an dual channel against dual channel.
It is very easy to improve speed by just putting two memory buses in parallel.
In practice the non optimal performance of DDR + P4 (a la Via or Intel 845D) depends mostly on the lower theoretical speed of 1 DDR channel against 2 RDR channels, implementation and the uneven divisor of a 400 Mhz system bus and a 266 Mhz memory.
In my opinion, it would be more correct to compare nForce based solutions for the P4 with dual channel DDR (existing but unavailable on the market because of legal reasons) to the i850's and i860's. And even then there would be the divisor factor and we should consider that the P4 couldn't anyway take advantage of all the bandwith of dual DDR with its bus clocked at standard fequency.
As far as the Athlon platform is concerned, it can't get a big advantage of the dual channel DDR system of nForce because of its system bus bottleneck (64 bits by 266 Mhz). But here we are speaking of two different architectures the extra bandwidth in that case is used mostly for the needs of the integrated graphics adapter.
BTW, it would be interesting to see how much an Athlon system could take advantage of higher available bandwitdth. In my opinion Athlons are more conservative in bus usage in respect to P4's to achieve similar performance but in the end they will anyway benefit from it at one point or another.
What do you think about ?