Once youâ€™ve invested a few billion pounds in a fabrication plant and have a team of skilled engineers beavering away youâ€™ll find it relatively easy to design and build an x86 processor. As time goes by youâ€™ll have to develop the processor to reduce production costs, raise efficiency and to increase the speed of the product. Over the last twenty years weâ€™ve seen processor manufacturers go through a ritualised dance to keep moving things along. They improve the production process to raise efficiency and production yields, and every few months they raise the clock speed multiplier. If things get tough the engineers can help out by increasing the amount of Level 2 cache which runs at Front Side Bus speed, but this is undesirable as L2 cache is expensive and takes up a great many transistors.
The problem here is that the transistor count is directly proportional to the area of the processor core, and the cost of production is almost entirely related to the size of the core. In an ideal world the core is as small as possible, while production yields are high, and that way you can turn the silicon wafer into as many functioning cores as possible. While PC users get fixated on clock speeds, youâ€™ll find that processor manufacturers are far more interested in the move from 200mm diameter wafers to 300mm wafers as that more than doubles the number of cores that you can produce at one time, which reduces costs handsomely.
The real bonanza comes when the fabrication plant is ready to move to a new production process. This involves a massive investment in stepping machine and lithography equipment but the move from 0.25 micron (Pentium II and III Katmai) to 0.18 micron (Pentium III Coppermine and Pentium 4 Willamette) and on to 0.13 micron (Pentium 4 Northwood) has reaped huge returns. As the process reduces in size Intel has been able to fit more transistors into the same area, which makes the core smaller and cheaper. At the same time communication between the transistors speeds up as the distances are reduced, and operating voltage can be dropped as the potential for voltage leaks diminishes over these shorter distances. As a direct result, clock speeds can be raised, resulting in a processor that is cheaper to produce, but which can be sold at a higher price.
There are all sorts of side issues, such as the material that is used for the interconnects (aluminium or copper), changes in the silicon (low-K or strained silicon are terms that are often bandied around), and whether or not the manufacturer should squeeze a bit more cache into the space that has been freed up by the new process but in the main these are options that the company can consider. In short a new fabrication process is a joyous experience which offers the potential of massive profits - if it works.