Dual core is something AMD had been thinking about when they designed their CPU called Hammer (Athlon 64/Opteron).
They removed the major bottlenecks and put the memory controller on the CPU chip. In a Dual CPU Opteron platform each CPU can talk directly to the other CPU and it's memory controller through a Hypertransport link. This will be the case with Dual core, plus each core will have it's own L1 and L2 cache. By the way, Dual Core CPUs should be pin-compatible with Socket 940.
Intel will have a more difficult time as they continue to deal with their bottlenecking, with the memory controller on the Northbridge down on the motherboard and everything sharing the FSB to the CPU. They will have some issues to overcome for Dual Core. If they don't address those issues, I would not be surprised if they add more Cache to cover up the lack of efficency and performance again.
Some think they are headed that way because it is becoming harder to scale CPUs. Some look forward to this direction of a parallel design in the Industry. It will really help improve Workstations and Server performance and costs. It will help with Multitasking and multithreaded applications on the Desktop, although single threaded applications will not see any improvement.
While he has some good points, I disagree with his overall negative and narrow view. Most recently I have not cared for his editorials.
Parallel designs can be taken advantage of even before code is rewriten for them. It's true that most Desktop programs would see little gain compared to Workstation or Servers but programers can rewrite the code to get even bigger performance gains by writing their programs to take advantage of the parallel design.
He's right, a single CPU would be more efficent. Hyperthreading is also more efficent than Dual CPUs but Dual CPUs perform substantially better.
Parellel designs have been talked about, researched and even used (For Servers or Super Computers) for some time. Right now it is a path that looks the most promising for added performance as it is not something they have done. They have increased the speed of their CPUs and created new designs to work better and yet they still cannot execute more than one task at a time.