May 05, 2012, 7:45 AM — Moore's Law -- the idea that computing power doubles roughly once every 18 months -- has proved to be surprisingly accurate since it was first outlined by Gordon Moore, a co-founder of Intel.
MORE ON PROCESSORS: AMD gains x86 processor market share on Intel in Q1
However, in recent decades, high-profile pronouncements of its impending collapse have become nearly as regular. CCNY theoretical physics professor Michio Kaku has been predicting such a breakdown since at least 2003, and he reiterated his views in a recent video talk for BigThink. He said that the critical point will be reached within a decade.
Small, hot and leaky
The constant shrinking of transistors -- which is responsible for the increased density and consequently higher computing capacity of microprocessors -- is unsustainable, according to Kaku. Once transistors approach 5 nanometers in size, they will be subject to two key problems: High density will see running temperatures skyrocket to impractically high levels, and quantum mechanics suggests that the Heisenberg Uncertainty Principle will result in electron leakage from the chip.
It's worth noting that Intel's new Ivy Bridge chips -- which represent a 10-nanometer shrink from the previous generation -- have already been found to run noticeably hotter than their predecessors under overclocking, though this may be due in part to other structural factors. Nevertheless, this could suggest that transistor density and size are beginning to be a concern for microprocessors.
While three-dimensional chips -- also a feature of Ivy Bridge -- and parallel processing can potentially delay the collapse of Moore's Law, Kaku said that these workarounds will eventually reach their limits as well.
That said, the CCNY physicist asserted that new forms of computing may yet allow processing power to resume its speedy upward climb. Molecular transistors, which Kaku described as the use of molecules shaped like valves to represent binary states, hold a great deal of promise, but current fabrication techniques aren't up to the challenge of mass production. Quantum computers could eventually become still more powerful, but these are even less well understood.