Moore’s law is the observation that the number of transistors in a denseintegrated circuit doubles approximately every two years.
Despite a popular misconception, Moore is adamant that he did not predict adoubling “Every 18 months.” Rather, David House, an Intel colleague, hadfactored in the increasing performance of transistors to conclude that integratedcircuits would double in performance every 18 months.
Moore viewed his eponymous law as surprising and optimistic: “Moore’s law is aviolation of Murphy’s law. Everything gets better and better.” The observationwas even seen as a self-fulfilling prophecy.
Intel used chemical-mechanical polishing to enable additional layers of metalwires in 1990; higher transistor density via trench isolation, local polysilicon, and improved wafer yield.
Conditions on the third wire result in distinct conductive properties including theability of the transistor to act as a solid state memory.
In 2012, a research team at the University of New South Wales announced thedevelopment of the first working transistor consisting of a single atom placedprecisely in a silicon crystal.
In 2015, IBM demonstrated 7 nm node chips with silicon-germanium transistorsproduced using EUVL. The company believes this transistor density would befour times that of current 14 nm chips.
Chenming Hu, inventor of the FinFET The vast majority of current transistors onICs are composed principally of doped silicon and its alloys.
As silicon is fabricated into single nanometer transistors, short-channel effectsadversely change desired material properties of silicon as a functionaltransistor.
Below are several non-silicon substitutes in the fabrication of small nanometertransistors.
Compared to their silicon and germanium counterparts, InGaAs transistors aremore promising for future high-speed, low-power logic applications.
In 2009, Intel announced the development of 80-nanometer InGaAs quantumwell transistors.
Despite being double the size of leading pure silicon transistors at the time, thecompany reported that they performed equally as well while consuming lesspower.
In 2011, researchers at Intel demonstrated 3-D tri-gate InGaAs transistors withimproved leakage characteristics compared to traditional planar designs.
The company claims that their design achieved the best electrostatics of anyIII-V compound semiconductor transistor.
In 2012, a team in MIT’s Microsystems Technology Laboratories developed a 22nm transistor based on InGaAs which, at the time, was the smallest non-silicontransistor ever built.
As economist Richard G. Anderson notes, “Numerous studies have traced thecause of the productivity acceleration to technological innovations in theproduction of semiconductors that sharply reduced the prices of suchcomponents and of the products that contain them.” Intel transistor gate lengthtrend – transistor scaling has slowed down significantly at advanced nodes Whilephysical limits to transistor scaling such as source-to-drain leakage, limited gatemetals, and limited options for channel material have been reached, newavenues for continued scaling are open.
These increases are described empirically by Pollack’s Rule, which states thatperformance increases due to microarchitecture techniques are square root ofthe number of transistors or the area of a processor.
There are cases where a roughly 45% increase in processor transistors hastranslated to roughly 10-20% increase in processing power.
Moore wrote only about the density of components, “a component being atransistor, resistor, diode or capacitor,” at minimum cost.
Transistors per integrated circuit – The most popular formulation is of thedoubling of the number of transistors on integrated circuits every two years.
At the end of the 1970s, Moore’s law became known as the limit for the numberof transistors on the most complex chips.
As of 2016, the commercially available processor possessing the highestnumber of transistors is the 24 core Xeon Haswell-EX with over 5.7 billiontransistors.
It is not just about the density of transistors that can be achieved, but aboutthe density of transistors at which the cost per transistor is the lowest.
As more transistors are put on a chip, the cost to make each transistordecreases, but the chance that the chip will not work due to a defect increases.
In 1965, Moore examined the density of transistors at which cost is minimized, and observed that, as transistors were made smaller through advances inphotolithography, this number would increase at “a rate of roughly a factor oftwo per year“.
Combined with Moore’s law, performance per watt would grow at roughly thesame rate as transistor density, doubling every 1-2 years.
According to Dennard scaling transistor dimensions are scaled by 30% everytechnology generation, thus reducing their area by 50%. This reduces the delayby 30% and therefore increases operating frequency by about 40%. Finally, tokeep electric field constant, voltage is reduced by 30%, reducing energy by65% and power by 50%.[note 2] Therefore, in every technology generationtransistor density doubles, circuit becomes 40% faster, while powerconsumption stays the same.
The exponential processor transistor growth predicted by Moore does notalways translate into exponentially greater practical CPU performance.
Moore’s 1995 paper does not limit Moore’s law to strict linearity or to transistorcount, “The definition of ‘Moore’s Law‘ has come to refer to almost anythingrelated to the semiconductor industry that when plotted on semi-log paperapproximates a straight line. I hesitate to review its origins and by doing sorestrict its definition.” Hard disk drive areal density – A similar observation wasmade in 2005 for hard disk drive areal density.