As the processes have shrunk in size the fabrication plants have been experiencing some unusual physical effects. The main problem is a quantum issue called tunnelling. At school we are taught that atoms are like planets with electrons that orbit the nucleus like asteroids. Although this is a neat image it is quite inaccurate. At the quantum level atoms donâ€™t inhabit a fixed space but instead exist in a number of locations at once until they are fixed by being observed. It sounds like a plot line from Star Trek but itâ€™s an unfortunate fact of life that atoms leak through materials that appear to be solid, and it would seem that the big crunch came when Intel moved to the 0.09 micron Prescott Pentium 4.
Instead of seeing the usual gains of reduced power and raised speeds, Prescott actually worked slower than the previous Northwood Pentium 4. To add to the problem Intel had increased the length of the pipeline inside Prescott to make it easier to raise clock speeds in the future which should have been invisible to outside observers, but as Prescott was such a disappointment we all noticed the performance hit, and none of us could ignore the TDP (Thermal Design Power) which broke through the 100W mark.
As the core area shrinks it is essential to keep heat dissipation under control. If the amount of power remains constant as the area shrinks then the thermal loads rises proportionate to the area, but between Willamette and Prescott we have seen the core area halve while the TDP has doubled. The result is that Intel has to find a way to shed four times the heat from each square millimetre of core area, and thatâ€™s no mean feat when youâ€™re using a block of finned aluminium with a fan on the top, rather than water cooling or active refrigeration.
Intel responded swiftly by canning development of the 0.065 micron Tejas Pentium 4 and by introducing a Pentium 4 Extreme Edition based on its Gallatin core that was dressed up as a Pentium 4. This was a cynical effort to show that it could compete with the AMD Athlon 64, and although it worked to a certain extent we all knew that Intel had big problems. The third part of its response was to move its development effort behind dual core processors.
Part four of the plan meant that Intel followed AMDâ€™s lead by introducing a model numbering system to break the link between clock speed and the perception of performance. In many ways this makes a great deal of sense as the fact that both the 550 and 760 run at 3.4GHz on an 800MHz FSB doesnâ€™t really matter as they use different amounts of L2 and L3 cache so itâ€™s tricky to compare the two processors directly. At one point in the past Intel had four permutations of 600MHz Pentium III processors which used different amounts of cache and ran on different FSBs, so the fact that they shared a clock speed merely confused matters. It was clear that Intel had been clinging to clock speeds as a marketing tool with which it could beat AMD over the head but the time had come to move on.