Semiconductor scaling and concurrent clouds - Part II

II. The era of concurrency

In Part I we discussed the technical secrets of semiconductor scaling that has kept Moore's Law going to this day. We learnt about such things as channel length scaling, high-K, leakage current, finFET, and timing closure. In particular we saw that as channel length is aggressively scaled down, the dynamic power dissipation goes up as the CUBE of clock frequency.

We will now apply these semiconductor insights to understand the shift to multi-core commodity processors, and the emergence of concurrent computing on the cloud.