It is a common misconception. Athlon 64 X2 cores cannot directly communicate. This mistake is often made because they are said to share be tied together, where as the Pentium D cores are not. In fact, Pentium D Presler core is two completely physically separate Cedar Mill cores,
Athlon 64 X2 processors have to be attached because they depend on an integrated memory controller for access to RAM while the Pentium D relies on the Northbridge. Athlon 64 X2s are still just a couple of Opterons tied together, and can only exchange data through the Northbridge across the HyperTransport. This has been proven multiple times by real processor architecture experts. Some might even think that a 2 GHz HT is faster than a 800 MHz FSB by a long shot, but this is
not really true. The only consumer x86 CPU with processor intercommunication via L2 cache is the Core 2.
The reason the Athlon 64 X2 out performs the Pentium D, or more specifically while the Athlon 64 out performs the Pentium 4 in games and other linear benchmarks, is because of the basic design of the micro-architecture itself. Mostly, it's due to Athlon 64's relatively short 12 stage pipeline vs Pentium 4's 31 stage pipeline. (Prescott and later). The consequence of this is Athlon 64/X2 is a benchmark winner but Pentium 4/D has better content authoring, multitasking and general multimedia performance.
It is an unfortunate fact that most authors for PC magazines are writers first and expert second. It's rare to actually read a piece in one from someone who actually knows what they're talking about.