Intel has it right, and they are perfectly justified reserving 8-threadedness for the desktop i7's. Mainstream users have been presented with general purpose multiprocessing (x86/x64) scenarios for years and all of the ideal cases have long since passed to the IGP (multimedia encode/decode, rich web geometry). It was only relevant to have numerous strong cores in the 90-65nm timeframe because high-def flash video and web rendering could not yet be offloaded to the GPU (unless you were in the 1% with a high-end card). Netflix and their use of silverlight is one of the last exceptions to this, but eventually that domino will fall to the IGP as well.
We now live in a time when almost any multithreaded job within the scope of an everyday PC user would be done faster and with less power on the IGP. In fact, the prime domain of multicore consumer CPUs is multitasking, not multiprocessing, and to that end, the balance you want to give your customers is a couple of the fastest CPUs they can get, not a truckload of the smallest CPUs they can get, particularly when you sell more laptops than desktops and your TDP is 17-35 and not 95-125.
But even if intel didn't have its manufacturing or power advantage, they would probably win in the long run (and indeed Kentsfield was faster than Barcelona, when the technology field was far closer to level). As you add cores, the performance increase that you can get, even in an unachievably ideal hypothetical multithreading scenario, decreases rapidly, and that is how you can tell AMD's philosophy is failing, even though it may not have failed completely just yet. They really peaked at Thuban and then a concert of wrong choices were brought into Bulldozer, but the central point of it all is that heavy multiprocessing is not what consumers are doing, so the moment AMD decided to shoot for the aggregate throughput niche was also the moment that they abandoned/surrendered to intel the much larger and more important general computing/consumer/productivity space.
We now live in a time when almost any multithreaded job within the scope of an everyday PC user would be done faster and with less power on the IGP. In fact, the prime domain of multicore consumer CPUs is multitasking, not multiprocessing, and to that end, the balance you want to give your customers is a couple of the fastest CPUs they can get, not a truckload of the smallest CPUs they can get, particularly when you sell more laptops than desktops and your TDP is 17-35 and not 95-125.
But even if intel didn't have its manufacturing or power advantage, they would probably win in the long run (and indeed Kentsfield was faster than Barcelona, when the technology field was far closer to level). As you add cores, the performance increase that you can get, even in an unachievably ideal hypothetical multithreading scenario, decreases rapidly, and that is how you can tell AMD's philosophy is failing, even though it may not have failed completely just yet. They really peaked at Thuban and then a concert of wrong choices were brought into Bulldozer, but the central point of it all is that heavy multiprocessing is not what consumers are doing, so the moment AMD decided to shoot for the aggregate throughput niche was also the moment that they abandoned/surrendered to intel the much larger and more important general computing/consumer/productivity space.