They've done 4 core in the same way that they did 2 core with the P4s, they've taken two two core dies and put them inside the same package. They have no dedicated interconnect between the two pairs of cores, so the front side bus is going to be a bottleneck. AMD's 4x4 might have its 4 cores in two separate packages BUT they're going to each have their own memory (meaning 2x memory bandwidth) AND dedicated interconnect too. Intel's may be faster for a couple of threads, but AMD's setup will scale better when you're using lots of interconnect or memory accesses!