This doesn't make sense, why is AMD even considering it if only the best DDR2 offers an advantage? Last I checked they weren't exactly losing the performance battle with intel, and with intel stuck at 3.8GHz single core, and AMD moving up, I don't think memory bandwidth is of concern here, at least not with single core (then there's the evidence that low latency has more of an effect than high bandwidth, although both is nice, you certainly don't get low latency with DDR2). Perhaps dual core needs more bandwidth? Or will need more bandwidth? I'd think they'd need to tweak the memory controller quite a bit to take a true advantage. I guess if they can can get extra performance and drive the nail even further in to the coffin that is the Prescott core...
But from a strategy standpoint I'm not sure it makes the best sense, forcing Intel to two failed memory types would be too good to be true, especially when DDR1 keeps giving us suprises as shown off by some of the best modules out there. I'd say leave intel to explain why they forced their users onto new platforms on to have them die almost as fast as they were introduced. DDR1 has been around for so long and is still alive AND kicking. DDR2 is doomed to a short life, perhaps the only way it would make sense to me would be in an FX type situation where the AMD flagship useses the new stuff, but that would mean even more specialization and just wouldn't be as cost effective.