The End of Moore's Law (AKA the Wall)

Moore's Law has so far proved remarkably accurate, and could in principle remain until 2015 before it is actually confronted with quantum effects. In 2015, processors should contain more than 15 billion transistors.[1]

Current Technology

In 1999 the CMOS transistor was pushed to its limits to the principle of MOSFET with a section of 18 nanometers (the size of about 70 atoms placed side by side) this is 7 times smaller than the smallest industrial transistor in 2003 (130 nm in 2003, 90 nm in 2004, 65 nm end 2005). It would allow a degree of theoretically 7 billion junctions on a coin the size of a quarter.[1] But they did it only as a simple search experiment to study the operation of CMOS technology when approaching molecular sizes. It would still be unable so far to control the coordinated assembly of a large number of transistors on a circuit and even less to do this on an industrial plan.

The cost of the technology to achieve chips incorporating more and more transistors increases in dizzying proportions. Another empirical law in Silicon Valley, states that the cost of manufacturing a chip doubles every four years since the manufacturing process, photolithography, used the past four years grows ever closer to its physical limits. Thus in 2004, Intel announced an investment of over $ 2 billion in its plant in Arizona Fab12 to manufacture chips from wafers 300 mm in diameter, which replaced the 200mm wafers in late 2005.[3]

The Wall

To make ever more tiny transistors, small radiation wavelengths are being used and the race to lead miniaturization photolithography in the X-ray spectrum is growing increasingly hard.[2] But in this range of wavelengths it becomes difficult or impossible to effectively concentrate the rays. In the mid-1990s, it was felt not to be able to manufacture industrial transistors under 400 atoms (100 nm) section with such a process.[1] Today, it is possible to push the critical dimensions of CMOS transistors to about 20 nm, lab demonstrations have already taken place. But these final orders will then hit the dimensional limits for physical and industrial technology. In the silicon industry, this limit is called the Wall.

References:

[1] Wu-Chun Feng (October 2003). "Making a case for Efficient Supercomputing"

[2] Understanding Moore's Law: Four Decades of Innovation. Edited by David C. Brock

[3] Manek Dubash (2005). "Moore's Law is dead, says Gordon Moore"