Once IBM began using integrated circuits to build computers, the development was market driven. Economic incentives encourage producers to crowd more and more components on each integrated circuit. If you double the number of components, the speed of an integrated circuit approximately doubles because the electronic components are closer together and smaller transistors have a shorter delay in switching. At the same time the variable production cost increases by less than two because while the details on the mask become smaller and smaller, the number of masking, etching and building steps stays constant. Nevertheless, as the number of components increases, the fixed costs in the form of capital costs to build new fabrication plants increases. The empirical rate of growth of electronic components on an integrated circuit is known as Moore's Law. From 1960 to 1972 the number of components on an integrated circuit doubled each year. From 1972 to the present and perhaps until the year 2000 the number of components on an integrated circuit will double every 18 months. In l977 it was possible to place 250,000 electronic components on an integrated circuit and by the year 2000 it may be possible to place 1 billion components on an integrated circuit.
The use of electronic components has increased 2000 times from 1960 to 1977 and is still increasing. A rule of thumb is that costs fall 20 to 30%with each doubling of output. For example, the production cost of the four-function electronic calculator fell from $100 to $5 while at the same time, the general price level of all goods more than doubled. As a student, you need to distinguish between (1) the decline in cost when the number of components is doubled and (2) the decline in cost when output is doubled. For example, when technology moved from a 1M memory to the 4M memory, the number of components increased by 4, but the cost of producing 4M memory integrated circuits was less than 4 times the cost of producing 1M memory integrated circuits. When the output of 4M devices increased by 2, the cost of 4M devices fell by 20 to 30% as efficiencies in production were obtained. Consequently, the fall in cost of a bit of memory has been greater than 30% per year since 1960. This falling cost can be attributed to both cost effects.
The physical limit to the number of components per integrated circuit is temporarily limited by the resolution distance of light. Using X or Gamma rays it is physically possible to pack even more components per integrated circuit. Ultimately, the physical limit is determined by the number of atoms required to construct a circuit. But before that limit is reached the physics of ICs will shift from classical to quantum mechanics.
The economic limit is determined by the success rate in production. As more and more components are packed into the integrated circuit the failure rate goes up. One cause of such increasing failures is dust particles in the air. To reduce the possibility of contamination, integrated circuits are made in specially constructed clean rooms which filter out dust. To reduce the dust even more, people are being removed from the production process through automation. Currently, and for the foreseeable future, the economic limit rather than the physical limit will control the number of components which can be placed on an integrated circuit. Achieving economic success rates as the number of components increases is why the fixed costs of creating new IC production facilities are rapidly increasing. Over time this factor will definitely slow down or even reverse the downward trend in the costs of electronic components.
Two other economic factors in IC design and production are the fixed cost of building a production foundry and the amount of heat generated by ICs. As the number of components on an IC has increased so has the construction cost of a new production facility. Now the cost is billions of dollars so that even major players now sometimes construct such facilities as joint ventures. The amount of heat generated by ICs has also rapidly increased with increasing numbers of components. A 2Bhz pentium 4 generates 70 watts of heat and a supercomputer as much as 5000 watts. Cooling computers is becoming a major problem. Designers are searching for ways to reduce heat production of ICs.
The end of Moore's law may occur in the second decade of the 21st century when IC makers hit insurmountable barriers and have to shift to quantum mechanics ICs, nanotubes, or building ICs atom by atom-Nanotechnology. Progress will probably continue with combinations of new approaches with current IC technology.
Moore's Law
Today worldwide commercial competition exists between US, Japanese, Korean, and European companies and this competition among firms is what made Moores Law an empirical fact . The US market was created by Silicon Valley, CA startup firms, which still excel in new product innovation. The US developed silicon compilers, which are software programs, to design custom integrated circuits automatically. Japan's strength is production efficiency of commodity integrated circuits such as memory chips. The Japanese tried to use memory production efficiencies as the means to dominate the entire microelectronics market. To gain a strategic hold on the memory market, Japanese firms in the 80s sold memory integrated circuits below cost until most of all but two US firms abandoned the memory market. IBM, not to become a captive of the Japanese firms, creates its own memory chips. IBM has kept pace in memory chips with the Japanese. Japan gained the lead in IC production.
To compete a firm must simultaneously crowd more and more electronic components into a chip and at the same time constantly improve the efficiency of the production process. In order to improve their competitive position in relationship with the Japanese, the US set up a research facility for improved integrated circuit production here in Austin called Sematech. The production efficiency of ICs is dependent on the size of the wafer on which ICs are made. The bigger the wafer the more efficient the process. Competition has created incentives for shifting production to larger and larger wafers. US and Korea got back into the game when Japanese firms hesitated in investing in the 300mm wafer size. The IC indusry has now moved to 300mm waferss and the US firms concentrated their research efforts in more profitable types of ICs such as microprocessors, communication ICs and specialty ICs. Recently the US electronics industry has regained the worldwide lead in microelectronics sales and now Samsung of South Korea is the number one manufacturer of memory chips.