Moore's Law means that the number of transistors that can be accommodated on an IC will double approximately every 18 months, and the performance will also double. Moore's Law was discovered by Gordon Moore, chairman emeritus of Intel, after long-term observation.
The first law of computers - Moore's Law Moore's Law In 1965, Gordon Moore prepared a report on the development trends of computer memory. He compiled a list of observations. When he started plotting the data, he discovered a surprising trend. Each new chip contains roughly twice the capacity of its predecessor, and each chip is produced within 18-24 months of the previous one. If this trend continues, computing power will increase exponentially relative to time periods. Moore's observations, now known as Moore's Law, describe trends that continue to this day and remain unusually accurate. This was also found to apply not only to describing memory chips, but also to accurately describe the development of processor power and disk drive storage capacity. This law forms the basis for performance predictions in many industries. In 26 years, the number of transistors on a chip increased more than 3,200 times, from 2,300 in the first 4004 launched in 1971 to 7.5 million in the Pentium II processor.
Due to the uniqueness of high-purity silicon, the higher the integration level, the cheaper the transistor becomes, which also leads to the economic benefits of Moore's Law. In the early 1960s, a transistor cost 10 US dollars. But as transistors get smaller and smaller, until 1,000 transistors can be placed on a hair, the price of each transistor is only one thousandth of a cent. According to relevant statistics, based on the price of 100,000 multiplications, the IBM 704 computer costs 1 US dollar, the IBM 709 drops to 20 cents, and the IBM 360 system computer developed by IBM at a cost of 5 billion in the mid-1960s has dropped to 3.5 cents. What exactly is "Moore's Law'"? To sum up, there are mainly three "versions" as follows:
1. The number of circuits integrated on an integrated circuit chip doubles every 18 months.
2. The performance of microprocessors doubles every 18 months, while the price drops by half.
3. The computer performance that one dollar can buy quadruples every 18 months.
Among the above statements, the first statement is the most common. The second and third statements involve price factors, and their essence is the same. Although the three statements have their own merits, they are identical on one point, that is, the "doubling" cycle is 18 months. As for "doubling" (or quadrupling), it is "the information integrated on the integrated circuit chip. "Number of circuits", "performance of the entire computer", or "performance that a dollar can buy" is a matter of opinion.
[Edit this paragraph] The origin of Moore's Law
The founder of "Moore's Law" is Gordon Moore, one of the founders of the famous chip manufacturer Intel. The rapid development of the semiconductor manufacturing industry from the late 1950s to the early 2000s led to the introduction of "Moore's Law".
As early as 1959, Fairchild, a famous American semiconductor manufacturer, first launched planar transistors, followed by planar integrated circuits in 1961. This planar manufacturing process uses a so-called "photolithography" technology to form components of semiconductor circuits, such as diodes, transistors, resistors and capacitors, on very flat silicon wafers. As long as the accuracy of "photolithography" continues to improve, the density of components will also increase accordingly, thus having great development potential. Therefore, planar technology is considered "the key to the entire semiconductor industry" and is also the technical basis for the emergence of Moore's Law.
On April 19, 1965, Moore, then director of the Fairchild Semiconductor Research and Development Laboratory, was invited to write an observation and review report for the 35th anniversary special issue of "Electronics" magazine, titled: "Let Integrated circuits are packed with more components.” Moore was asked by the magazine to predict the development trends of the semiconductor component industry over the next decade. According to his calculations, by 1975, it would be possible to have 65,000 components on a single silicon chip with an area of ??only a quarter of an inch square.
He made this inference based on the linear relationship between device complexity (circuit density increases and price decreases) and time. His exact words are as follows: "The complexity at the lowest component price increases by about one per year. times. It is certain that this growth rate will continue if not accelerate in the short term, while the growth rate in the longer term should be slightly volatile, although there are good reasons for this growth rate. It will remain almost constant for at least the next decade." This was the original prototype of what became known as "Moore's Law."
[Edit this paragraph] Modification of Moore's Law
In 1975; Moore submitted a paper at the International Telecommunication Union IEEE's annual academic meeting. Based on the actual situation at the time, " The growth rate of "density doubled every year" has been re-examined and revised. According to Moore himself in an interview with a (Scientific American) editor in September 1997, he changed "double every year" to "double every two years" and stated that he never said that "Doubling every 18 months".
However, according to some online media, shortly after Moore's paper was published, someone revised his prediction to "The density or capacity of semiconductor integrated circuits will double every 18 months, or double every 18 months." "Grown 4 times in three years", someone even listed the following mathematical formula: (circuit growth rate per chip) = 2 (year - 1975) / 1.5. This statement later became the "unknown knowledge" of many people and has been passed down to this day. Moore's own voice, whether it was the original "double every year" or the later revised "double every two years", was drowned out and is now rarely heard.
History has actually made a big joke with people: It turns out that the currently widely circulated "Moore's Law" is not Moore's own statement!
[Edit this paragraph] Verification of Moore's Law
Is Moore's Law accurate? Let's look at some specific data first. In 1975, a new charge-front-load device memory chip indeed contained nearly 65,000 components, which was surprisingly consistent with Moore's prediction ten years ago! According to statistics released by Intel, the number of transistors on a single chip increased from 2,300 on the 4004 processor in 1971 to 7.5 million on the Pentium II processor in 1997, an increase of 3,200 times in 26 years. We might as well conduct a simple verification on this: If we follow Wal-Mart's prediction of "doubling every two years", 26 years should include 13 doubling cycles. After each cycle, the number of components integrated on the chip should increase by 2n times. (0≤n≤12), so by the 13th cycle, that is, after 26 years, the number of components should have increased by 212=4096 times. As a prediction of a development trend, this is quite close to the actual growth rate of 3200 times. If we take the 18-month doubling cycle mentioned by others, the two are far from each other. It can be seen that in the long run, Moore's own statement is closer to reality.
Some people also examine the correctness of Moore's Law from the three major elements of a personal computer (PC) - microprocessor chip, semiconductor memory and system software. In terms of microprocessors, from 8086 and 8088 in 1979, to 80286 in 1982, 80386 in 1985, 80486 in 1989, Pentium in 1993, Pentium Pro in 1996, and Pentium II in 1997, the functions are becoming more and more powerful. Prices are getting lower and lower, and every update is a direct result of Moore's Law. At the same time, the internal memory capacity of PCs has expanded from the earliest 480k to 8M and 16M, which is more consistent with Moore's Law. In terms of system software, due to the limitation of storage capacity of early computers, the scale and function of system software were greatly restricted. As memory capacity increases exponentially at the rate of Moore's Law, system software is no longer limited to a small space. It contains The number of lines of program code has also increased dramatically: Basic's source code only had 4,000 lines in 1975, and grew to about 500,000 lines 20 years later. Microsoft's word processing software Word, the first version in 1982 contained 27,000 lines of code, which increased to approximately 2 million lines of code 20 years later. Someone plotted its development speed on a curve and found that the size and complexity of software are growing faster than Moore's Law.
Developments in system software, in turn, increased demand for processors and memory chips, spurring faster development of integrated circuits.
What needs to be pointed out here is that Moore’s Law is not a law of mathematics or physics, but an analysis and prediction of development trends. Therefore, both its written expression and quantitative calculations should allow a certain amount of deviation. Margin. In this sense, Moore's prediction is indeed quite accurate and valuable, so it has been recognized by people in the industry and has had a huge response.
[Edit this paragraph] Application of Moore's Law
The year 2005 is the 40th anniversary of Gordon Moore, one of the founders of Intel Corporation, who proposed the famous "Moore's Law". In the past 40 years, the integration trend of semiconductor chips has been as predicted by Moore, promoting the development of the entire information technology industry, and thus bringing changes to the lives of thousands of households.
In April 1965, Moore, who was an electronics engineer at Fairchild at the time, published an article in Electronics magazine predicting that the number of transistors and resistors integrated on semiconductor chips would double every year. In 1975, he proposed a revision, saying that the number of transistors integrated on a chip would double every two years.
At that time, integrated circuits had only been around for 6 years. Moore's lab could only integrate 50 transistors and resistors on one chip. Moore's prediction at the time sounded like science fiction; since then, technical experts have continued to believe that the speed of chip integration has "reached its peak." But it turns out that Moore's prediction was accurate. Although the cycle of technological advancement has lengthened from the originally predicted 12 months to nearly 18 months today, "Moore's Law" remains valid. The most advanced integrated circuits currently contain 1.7 billion transistors.
"Moore's Law" summarizes the speed of progress in information technology. In the past 40 years, computers have transformed from mysterious and inaccessible behemoths into indispensable tools for most people. Information technology has entered countless ordinary households from laboratories. The Internet has connected the whole world, and multimedia audio-visual equipment has enriched everyone's life. .
The driving force behind all this is semiconductor chips. If transistors, resistors, and capacitors were installed on circuit boards in the old way, not only would personal computers and mobile communications not be available, but new technologies such as genomic research and computer-aided design and manufacturing would be impossible.
"Moore's Law" has also driven fierce competition in the chip industry. In commemorating the 40th anniversary of the publication of this law, Moore, who is the honorary chairman of Intel Corporation, said: "If you expect to be a leader in the semiconductor industry, you cannot afford to fall behind Moore's Law." From Fairchild in the past to today Intel, Motorola, Advanced Micro Devices, etc., the competition in the semiconductor industry around "Moore's Law" is as fierce as the waves rushing through the sand.
There is no doubt that "Moore's Law" has far-reaching significance for the entire world. When reviewing the progress of the semiconductor chip industry over the past 40 years and looking forward to its future, information technology experts say that "Moore's Law" may still apply in the next few years. But as transistor circuits approach their performance limits, this law will eventually come to an end. When will "Moore's Law" expire? Experts have different opinions on this.
Stan Williams, a researcher at HP Labs in the United States, said that by around 2010, semiconductor transistors may have problems, and chip manufacturers must consider alternative products. Paul Gargini, director of Intel's technology strategy department, believes that around 2015, some "hybrid" transistors using technologies such as nanowires will be put into production and will replace semiconductor transistors within five years. Some experts have pointed out that semiconductor transistors can continue to develop until the limit of their size - between 4 and 6 nanometers, which may be 2023.
Experts predict that as the size of semiconductor transistors approaches the nanoscale, not only side effects such as chip heating will gradually appear, but also the operation of electrons will be difficult to control, and semiconductor transistors will no longer be reliable. "Moore's Law" will certainly not continue to be valid for the next 40 years. However, new developments such as nanomaterials and phase change materials have emerged and are expected to be applied to future chips. By then, even if "Moore's Law" comes to an end, the pace of advancement of information technology will not slow down.
[Edit this paragraph] The evolution of Moore's Law
The great reputation of Moore's Law has caused many people to imitate its expression, thus deriving and multiplying multiple versions of "Moore's Law" ", such as:
Moore's Second Law: In the 30 years since Moore's Law was proposed, the performance of integrated circuit chips has indeed been greatly improved; but on the other hand, Intel executives began to notice that chip manufacturers The cost is also increasing accordingly. In 1995, Intel Chairman Robert Noyce foresaw that Moore's Law would be constrained by economic factors. In the same year, Moore wrote in the Economist magazine: "What worries me most now is the increase in costs... This is another exponential curve." His statement is known as Moore's second law.
New Moore's Law: In recent years, the term "New Moore's Law" has appeared in domestic IT professional media, which refers to the increasing rate of the number of Internet hosts and the number of Internet users in my country, approximately every Doubled in half a year! And experts predict that this trend will continue in the next few years.
[Edit this paragraph] Prospects of Moore’s Law
Moore’s Law has been around for 40 years. People are surprised to see the level of semiconductor chip manufacturing technology improving at a dizzying speed. Currently, the main frequency of Intel's microprocessor chip Pentium 4 has reached 2G (i.e. 12000M). In 2011, a chip containing 1 billion transistors and capable of executing 100 billion instructions per second will be launched. People can't help but ask: Will this incredible speed of development continue indefinitely?
You don’t need complicated logical reasoning to know that the geometric size of the components on the chip cannot be reduced indefinitely. This means that one day, the number of components that can be integrated per unit area of ??the chip will be reduced. will reach its limit. The question is simply what that limit is and when it will be reached. Experts in the industry predict that the growth rate of chip performance will slow down in the next few years. It is generally believed that Moore's Law can be applied for another 10 years or so. The restrictive factors are firstly technology and secondly economics.
From a technical perspective, as the density of circuits on silicon wafers increases, their complexity and error rates will increase exponentially, making comprehensive and thorough chip testing almost impossible. Once the width of the lines on the chip reaches the order of nanometers (10-9 meters), which is equivalent to the size of only a few molecules, the physical and chemical properties of the material will undergo qualitative changes, causing semiconductor devices using current processes to fail. If things work normally, Moore's Law is about to come to an end.
From an economic point of view, as stated in Moore's Second Law above, it currently costs 2-3 billion US dollars to build a chip factory. When the line size is reduced to 0.1 micron, it will soar to 10 billion US dollars. The investment in a nuclear power plant is huge. Unable to afford the money, more and more companies are forced to exit the chip industry. It seems that it will not be easy for Moore's Law to maintain its lifespan for another ten years.
However, some people look at the problem from a different perspective. Dan Linqi, president and CEO of an American company called CyberCash, said, "Moore's Law is a law about human creativity, not a law of physics." People who hold similar views also believe that Moore's Law is actually a law about human belief. When people believe that something can be done, they will work hard to achieve it. When Moore first presented his observation report, he actually gave people a belief that the development trend he predicted would continue.
Moore's Law is a conclusion drawn by Gordon Moore, honorary chairman of Intel Corporation, after long-term observations. It was first used to describe a phenomenon in the field of semiconductor manufacturing, that is, it refers to the amount of energy that can be accommodated on an integrated circuit. The number of transistors will double approximately every 18 months, and the performance will also double. Later, Moore's Law was introduced to other high-tech industries to describe the performance improvements brought about by the rapid development of technology.
In the optical fiber communications industry, Dense Wavelength Division Multiplexing (DWDM) technology once well explained Moore's Law. DWDM is a key basic network technology that significantly reduces bandwidth costs by transmitting multiple parallel gigabit optical signals within an optical fiber, thus making broadband Internet popular. This technology also has the advantages of long transmission distance and low latency.
With the rapid growth of network transmission volume, telecom operators hope to transmit more information at a lower unit cost. Therefore, DWDM's position in fixed communication infrastructure has been consolidated and continuously strengthened. From 2003 to 2007, operators' spending on DWDM technology nearly tripled. In 2007, global spending on this technology equipment reached US$5.8 billion.
For the past 10 years, the well-known consulting firm Ovum has used a measurement formula of network bandwidth capital expenditure (capex) to calculate the time required to transmit 1GB of information per second over a distance of one kilometer. cost. Telecom operators initially achieved a transmission rate of 64kbps on each voice line. Later, after each user used tens of thousands of megabytes of information, the voice lines were overwhelmed. Fortunately, fiber optic technology appeared. Competition among equipment suppliers has caused transmission costs to drop sharply. In 1993, before the emergence of DWDM technology, the cost of transmitting 1GB of information per second over a distance of one kilometer was US$2,000. By 2007, this number was less than 1 The U.S. dollar is developing at a rate that eclipses Moore's Law.
DWDM technology has been officially deployed for 13 years. However, it seems to have stopped the exponential growth that once created miracles in the history of telecommunications development and has entered the depression of adolescence. Analysts pointed out that if the entire system does not expand exponentially within the next five years, the geometric growth of DWDM will be difficult to maintain. Is the influence of Moore's Law over in the optical fiber communications market?
In the past years, the success of DWDM has relied on diverse innovations, such as fiber amplifiers and optical add-drop multiplexers (OADM), as well as lasers, wavemeters, filters, etc. Advances in technology and innovations in various system software have allowed the system to obtain higher capacity and improve operational flexibility.
Brutal competition has led to the continuous reduction of the cost of 10G networks and forced DWDM to continuously improve cost performance in 13 years, although during this period, investment in R&D of optical fiber systems and investment in component innovation were relatively low. Meanwhile, the entire industry was healing from the collapse of the telecoms bubble in the first few years of the decade, which caused spending to plummet in major markets. The consequence of cuts in R&D investment during that period was delays in the deployment and commercialization of more cost-effective 40G technologies.
However, the market finally chose 40G technology. The most likely to achieve exponential growth in transmission cost-effectiveness in the near future is 40G network technology. This technology provides four times the bandwidth of the existing 10G network, while the capex is only one-fourth, and the performance is not inferior at all. Although Ovum Consulting believes cost reductions of this magnitude are unlikely to be achieved before 2012, given the surge in investment in the technology over the past few years, miracles are still possible.
40G technological innovations that have been launched and will be launched in 2008 include: technology developed by Nortel Networks, which provides the best performance on the market and has a clear path to evolve to 100G technology; Opvista technology, It has obvious advantages in promoting the use of 40G technology in metropolitan area networks; standardized 40G module technology jointly developed by Stratalight, Mintera and other companies has also made progress. Optical fiber technology supplier Infinera is also working hard to solve the contradiction between cost, capacity and transmission distance through 40G technology innovation, and is expected to release new technologies later this year or next year.
At the same time, network operators and equipment suppliers will also promote the innovation of 100G technology, which will continue the success of DWDM and satisfy the increasing desire of global users for communication services.