The fact that Moore`s Law is approaching its natural death is perhaps most painfully present among chipmakers themselves; As these companies face the task of building ever more powerful chips against the reality of physical opportunities. Even Intel is competing with itself and its industry to create what might ultimately not be possible. According to Moore`s observation, one of the main attractions of embedded electronics is its low cost. This advantage increases as technology advances. A single semiconductor substrate can produce more complex circuit functions. The cost per component in simple circuits is almost inversely proportional to the number of components. However, the cost per component tends to increase as more components are added, and lower efficiencies offset the increase in complexity. As a result, costs are minimal at any point in technological development. For example, manufacturing cost forecasts per component in 1970 were only one-tenth of what they were in 1965.
Our editors will review what you have submitted and decide if the article needs to be revised. The ubiquitous and inexpensive computing that is rapidly spreading around us is fundamentally changing the way we work, play and communicate. The fundamental power of Moore`s Law has led to breakthroughs in modern cities, transportation, healthcare, education, and energy production. In fact, it`s pretty hard to imagine what our modern world could look like without Moore`s Law. Nevertheless, technologists have internalized Moore`s Law and become accustomed to believing that computer speed doubles every 18 months, as Moore observed more than 50 years ago, and until recently this was true. However, Moore`s Law is becoming obsolete. What for? And what alternatives do we have? In 1974, Robert H. Dennard of IBM recognized the rapid scaling technology of MOSFETs and formulated the so-called Dennard scaling, which describes their power density remaining constant as MOS transistors become smaller, so that power consumption remains proportional to the area. [17] [18] The scaling and miniaturization of MOSFETs were the main driving forces behind Moore`s Law.
[19] Semiconductor industry data show that this inverse relationship between power density and surface density broke in the mid-2000s. [20] But we could already be there. Robert Colwell, head of the microsystems technology office at the Defense Advanced Research Projects Agency, is using the year 2020 and 7nm as the last process technology node. «In reality, I expect the industry to do whatever it takes to move to 5nm, even if 5nm doesn`t offer much of an advantage over 7nm, and that pushes the end back to 2022 earliest. I think the ending comes right around those nodes. In 2016, the International Technology Roadmap for Semiconductors, after using Moore`s Law to move the industry forward since 1998, created its final roadmap. He no longer focused his research and development plan on Moore`s Law. Instead, he described what might be called the More than Moore strategy, where application requirements drive chip development rather than focusing on scaling semiconductors. App drivers range from smartphones to AI to data centers. [123] The slowing down of Moore`s Law has led many people to ask, «Is Moore`s Law dead? In fact, this is not the case.
Although Moore`s Law still brings exponential improvements, results are delivered more slowly. However, the pace of technological innovation is not slowing down. On the contrary, the explosion of hyperconnectivity, big data, and AI applications has accelerated the pace of innovation and the need for Moore`s Law-style improvements to the technology provided. This is the wording of Moore`s 1965 paper. [1] It is not only the density of transistors that can be achieved, but the density of transistors where the cost per transistor is the lowest. [152] The more transistors placed on a chip, the lower the cost of manufacturing each transistor, but the likelihood that the chip will not work due to a defect increases. In 1965, Moore studied transistor density, where costs are minimized, and observed that if transistors were made smaller by advances in photolithography, this number would increase at «a rate of about a factor of two per year.» [1] Several measures of digital technology are improving exponentially in the context of Moore`s Law, including component size, cost, density, and speed. Moore wrote only about component density, «a component is a transistor, resistor, diode, or capacitor,»[132] at minimal cost. For perspective, a nanometer is one billionth of a meter, smaller than the wavelength of visible light. The diameter of an atom varies from about 0.1 to 0.5 nanometers. Moore`s observation transformed IT from a rare and expensive enterprise into a ubiquitous and affordable necessity. All the modern computer technology we know and appreciate was born from the foundation laid by Moore`s Law.
From the internet itself to social media to modern data analytics, all of these innovations come directly from Moore and his ideas. This new design trend is driven by systemic complexity. Some aspects of this new design approach have been described as «more than Moore». This term mainly refers to 2.5D and 3D integration techniques. The observation is named after Gordon Moore, the co-founder of Fairchild Semiconductor and Intel (and former CEO of the latter), who postulated a doubling of the number of components per IC per year in 1965 and predicted that this growth rate would continue for at least another decade. In 1975, in preparation for the next decade, it revised forecasts to double every two years, a compound annual growth rate (CAGR) of 41%. Although Moore did not use empirical evidence to predict that the historical trend would continue, his prediction has been valid since 1975 and has since been known as the «law.» That same year, Mohamed Atalla and Dawon Kahng invented the MOSFET (metal-oxide-semiconductor field-effect transistor), also known as the MOS transistor, at Bell Labs. [5] The MOSFET was the first truly compact transistor that could be miniaturized and mass-produced for a variety of applications,[6] with its high scalability[7] and low power consumption, resulting in higher transistor density[8] and enabling the construction of high-density IC chips. [9] In the early 1960s, Gordon E. Moore realized that the ideal electrical and scaling properties of MOSFETs would lead to an increasing level of integration and unprecedented growth in electronic applications. [10] This is a controversial issue.
Although the evidence of the law is slowing, some still hope that there will be new applications that will perpetuate the legacy of the law. Another factor slowly killing Moore`s Law is the rising cost of energy, cooling, and manufacturing. Building new CPUs or GPUs (Graphics Processing Units) can be expensive. The cost of manufacturing a new 10nm chip is about $170 million, nearly $300 million for a 7nm chip, and more than $500 million for a 5nm chip. These numbers can only increase with a few specialized chips. For example, NVidia spent more than $2 billion on research and development to produce a GPU designed to accelerate AI. Due to the intensive footprint of resources and toxic materials used in computer manufacturing, obsolescence leads to serious adverse environmental impacts. Americans throw away 400,000 cell phones every day,[142] but this high level of obsolescence appears to companies as an opportunity to generate regular sales of expensive new devices rather than keeping a device for an extended period of time, leading the industry to use planned obsolescence as a profit center. [143] In the decades following Gordon Moore`s initial observation, Moore`s Law guided the semiconductor industry in long-term planning and setting research and development (R&D) goals. Moore`s Law has been a driving force behind the technological and social change, productivity and economic growth that characterize the late twentieth and early twenty-first centuries. On the other hand, manufacturers add specialized processing units to handle functions such as graphics, video, and cryptography. For example, Intel`s parallel JavaScript extension adds support for not only multiple cores, but also other non-generic processing capabilities of their chips as part of the migration of client-side scripting to HTML5.
[147] If we continue to miniaturize chips, we will undoubtedly encounter Heisenberg`s uncertainty principle, which limits precision to the quantum level and therefore limits our computational capabilities.