There is a myth about a rancher who decides to build his strength by lifting a calf. From the day the calf is born, the rancher goes each day into the barn and raises the calf over his head. His reasoning is that if he lifted the calf yesterday, he should be able to lift the same animal today. Eventually, the rancher figures, the calf will be full-grown and the rancher will thus be strong enough to lift an adult cow.

Obviously, this reasoning is flawed. The day of reckoning will come when, despite the previous days’ successes, the rancher will no longer be able to lift the calf. The change would come gradually and the actual day of failure would be difficult to predict accurately, but it would be inevitable nonetheless. So it is with metrics that change gradually but continuously over a long period of time. Even though there is no apparent discontinuity, the day invariably comes when the old way is out and the new way is in.

In his paper published on April 19, 1965, Gordon Moore predicted that the number of transistors on a chip would approximately double every two years. Almost 40 years later, that revolutionary prediction remains remarkably close to reality. Although Mr. Moore’s forecast shows no convincing signs of weakness yet, there have been many days of reckoning along the way. SSI gave way to MSI, then LSI and VLSI. Semi-custom devices replaced standard parts for many applications. Gate arrays boomed onto the scene, almost went extinct during the cell-based ASIC age, then recently re-emerged in disguise as “structured ASIC” devices. Programmable logic arrived and grew explosively into one of the dominant implementation technologies available today. [more]