Moore's Law: Problematic Future Or Technological Promise?

Moore's Law, the prediction that the number of transistors in an integrated circuit would double every two years, has been a driving force behind technological and social change, as well as economic growth, for the past 50 years. However, the law is facing problems due to the physical limitations of miniaturization and the increasing costs of research and development. While some industry leaders argue that Moore's Law is already dead, others believe it will continue to be relevant for at least the next decade.

The speed of light and atomic nature of materials

The speed of light is a fundamental physical constant, currently accepted to be 299,792,458 metres per second, or about 186,282 miles per second. It is a universal physical constant, acting as the upper limit for the speed at which conventional matter or energy can travel through space.

The speed of light is finite, constant, and provides a natural limitation on the number of computations a single transistor can process. As information cannot be passed quicker than the speed of light, the speed of computation is limited by the speed of an electron moving through matter.

The speed of light is also significant because it is considered a fundamental constant of nature. It serves as the single limiting velocity in the universe, acting as an upper bound to the propagation speed of signals and to the speeds of all material particles.

The atomic nature of materials also poses a problem for Moore's Law. Transistors can only get so small, and eventually, the more permanent laws of physics get in the way. Transistors are now measured on an atomic scale, with the smallest commercially available being 3 nanometres wide, only slightly wider than a strand of human DNA.

As transistors continue to be miniaturised, Heisenberg's uncertainty principle will become a limiting factor, restricting precision at the quantum level and, therefore, limiting computational capabilities.

Growing costs

Moore's Law states that the number of components on a single chip doubles every two years at minimal cost. However, as transistors get smaller, the cost of manufacturing them increases. The cost to manufacture a 10nm chip is around $170 million, while a 5nm chip costs over $500 million. The cost of the tools used to manufacture chips also doubles every four years. This is leading to rising manufacturing costs, which is an important consideration for the future of Moore's Law.

The cost of building new CPUs or GPUs can be extremely high. For example, NVidia spent over $2 billion on research and development to produce a GPU designed to accelerate AI. The cost of a fab (a semiconductor fabrication plant) is also rising, at around 13% per year, and is expected to reach $16 billion or more by 2022. This has reduced the number of companies with plans to make the next generation of chips from 25 in 2002 to just three in 2022.

The increasing costs of manufacturing chips are causing concerns about the future of Moore's Law. Some believe that the physical limits of the law will be reached in the 2020s, as it is unlikely that transistors can be made smaller than atoms. There is also the issue of cooling to consider, as adding more components to a chip in a small space causes it to heat up.

The end of Moore's Law could have a significant impact on the development of technology. Moore's Law has been a driving force of technological and social change, as well as productivity and economic growth. It has influenced the progress of computing power and led to advancements in digital electronics, such as the improvement of sensors and the number of pixels in digital cameras.

The end of Moore's Law could also affect the cost of computers. As the cost of producing more advanced chips rises, this could lead to higher prices for consumers. It could also slow down the rate of improvement in computer performance. However, there are alternative ways to improve computer performance, such as through software performance engineering and improvements in software, algorithms, and hardware architecture.

The uncertainty principle

Moore's Law is based on an observation made by Gordon Moore in 1965. Moore noticed that the number of transistors in an integrated circuit (IC) had doubled every year between 1960 and 1965, and he projected this rate of growth to continue for at least another decade. In 1975, he revised his forecast, stating that the number of transistors in an IC would double every two years.

The prediction has held since 1975 and has become known as a "law". However, Moore's Law is not an actual law of physics but rather an empirical relationship. It is an experience-curve law that quantifies efficiency gains from experience in production.

The law states that the speed and capability of computers can be expected to double every two years as a result of increases in the number of transistors a microchip can contain. This has significant implications for technological progress and computing power.

However, Moore's Law is facing challenges due to the physical limitations of transistor size. Transistors are now measured on an atomic scale, and the laws of physics, such as the speed of light and the atomic nature of materials, impose natural limitations on transistor size and computational capabilities.

As transistors approach the size of atoms, Heisenberg's uncertainty principle comes into play, limiting precision at the quantum level and, thus, our computational capabilities. James R. Powell calculated that due to the uncertainty principle alone, Moore's Law will become obsolete by 2036.

Additionally, the cost of continuing to miniaturize transistors is skyrocketing, with the cost of manufacturing smaller chips running into the hundreds of millions of dollars.

Despite these challenges, some industry leaders, such as Intel CEO Pat Gelsinger, believe that Moore's Law is still relevant and can be upheld or even outpaced in the coming years. They argue that Moore's Law is still driving improvements in processing technology and enabling more intelligent edge devices and endpoints.

In conclusion, while Moore's Law is facing challenges due to physical limitations and increasing costs, its impact on technological progress over the past five decades cannot be understated. As we approach the limits of Moore's Law, alternative approaches such as quantum computing, specialized chip architecture, and improved software performance engineering will become increasingly important to sustain computational progress.

The cost of new CPUs or GPUs

The cost of developing new CPUs or GPUs is a significant factor in the obsolescence of Moore's Law. The cost of manufacturing a new 10nm chip is around $170 million, with the cost increasing to almost $300 million for a 7nm chip, and over $500 million for a 5nm chip. These costs are expected to increase as chips become smaller. For example, NVidia spent over $2 billion on research and development to produce a GPU designed to accelerate AI.

The increasing costs of energy, cooling, and manufacturing are also contributing factors. As transistors get smaller, wires and transistors become more resistant to the flow of current, and it becomes more difficult to perform correct computations. This leads to higher energy consumption and cooling requirements, further increasing costs.

The end of Moore's Law has significant implications for the future of computing. As the law predicted that the number of transistors in a dense integrated circuit would double every two years, the end of this trend means that the pace of technological progress may slow down. However, there are alternative technologies being developed that could replace Moore's Law, such as quantum computing, field-programmable gate arrays, spintronics, and optical computing.

In the short term, companies can prioritize education and up-to-date training for their programmers to help mitigate the increasing costs of computing power. Additionally, focusing on software performance engineering and using efficient coding methods can help improve computing performance without relying solely on hardware advancements.

The need for specialised chip architecture

Moore's Law, the prediction that the number of transistors in an integrated circuit would double about every two years, has been a driving force of technological and social change, productivity, and economic growth. However, as transistors approach the atomic level, the law is becoming obsolete. The end of Moore's Law means that the pace of progress will slow down, and there will be a need to find alternative ways to improve computer performance.

One solution is to focus on software performance engineering and develop more efficient systems. Another option is to transition to new chip architectures, such as specialised chip architecture. This approach is already gaining traction due to the high demand for machine learning applications.

Specialised chip architecture refers to the instruction set architecture (ISA), which is a detailed set of rules and specifications that govern how a chip processes data and executes instructions. The ISA defines the types of instructions the chip can understand, the format of data it can manipulate, and the registers used for temporary storage during computations.

There are two main categories of chip architecture: x86 Architecture, a complex instruction set computer (CISC) architecture developed by Intel, and RISC (Reduced Instruction Set Computing) Architectures, which focus on simpler instructions that can be executed faster and with lower power consumption.

The future of chip architecture may include heterogeneous integration, where various processing cores are integrated into a single package, each optimised for specific tasks. Another possibility is 3D chip stacking, where multiple layers of chips are stacked vertically to create 3D structures, allowing for more processing power and functionality within a smaller footprint.

The development of specialised chip architecture is an important area of innovation as it can help improve processing power and efficiency, even as Moore's Law becomes obsolete.

Frequently asked questions