If our electronics are to become smaller and faster, we need technological advances.
We live in a world driven by computer circuits. Modern life depends on transistors on semiconductor chips and silicon-based integrated circuits that can switch electronic signals on and off. Most transistors use a rich and inexpensive element of silicon because it either blocks or allows current to flow, and it is both an insulator and a semiconductor.
Until recently, the size of tiny transistors squeezed on silicon chips was reduced by half each year. It created the modern digital age, but this era is coming to an end. With the advent of compute-intensive tasks such as the Internet of Things, artificial intelligence, robotics, autonomous vehicles, 5G and 6G mobile phones, the future of technology is at stake. So what happens next?
What is Moore's Law?
Moore's Law is an exponential increase in computing power. As early as 1965, Intel co-founder Gordon Moore observed that the number of transistors on an inch of computer chip doubled every year, and the cost was halved. Now, this time is 18 months, and it is getting longer. In fact, Moore's Law is not a law, but the observations of a person working for a chip maker, but the increased time means that future intensive computing applications may be threatened.
Is Moore's Law dead?
No, but it's too slow and the silicon chip needs help. Stephen Doran, chief executive officer of Catapult, a British semiconductor application company, said: "In more and more applications that require speed, latency and light detection, silicon is reaching its performance limit."
However, he thinks it's too early to talk about silicon alternatives. He added: "This means that silicon will be completely replaced, which is unlikely to happen in the short term and will most likely never happen."
David Harold, vice president of marketing communications at Imagination Technologies, said: "At least until 2025, Moore's Law-like performance improvement still has potential. Until the 1940s, silicon will still dominate the chip market."
The second era of computers is coming
It's important to study the silicon transistor problem carefully; as a concept, it doesn't "die", but it has surpassed its peak. "Moore's Law specifically refers to the performance of integrated circuits made from semiconductors, and only records calculations over the past 50 years," said Craig Hampel, chief scientist of Rambus' Memory and Interfaces Division.
"The growing demand for computing in humans can be traced back to abacus, mechanical calculators, and vacuum tubes, and may extend far beyond semiconductors (such as silicon), including superconductors and quantum mechanics."
Beyond silicon is a problem, as future computing devices will need to be more powerful and flexible. Harold said: "The increasing computing problem is that future systems will need to learn and adapt to new information. They must be" like the brain. "Coupled with the transformation of chip manufacturing technology, they will create a revolutionary second for computing Times. "
What is cold computing?
Some researchers are working on new ways to obtain high-performance computers with less energy. "Cold operation of a data center or supercomputer can bring significant performance, power, and cost advantages," Hampel said.
Microsoft's Natick project is an example. As part of the project, a huge data center sank into the coast of the Orkney Islands in Scotland, but this is only a small step. Lowering the temperature further means less leakage current and a lower threshold voltage for transistor switching.
"It reduces some of the challenges of extending Moore's Law," Hampel said. He added that for these types of systems, the natural operating temperature is 77K (-270 ° C) liquid nitrogen. "The atmosphere is rich in nitrogen, it is relatively cheap to collect in liquid form, and it is an effective cooling medium. We hope that in terms of memory performance and power consumption, it may be extended for another 4 to 10 years."
What is a compound semiconductor?
The next generation of semiconductors consists of two or more elements whose properties make them faster and more efficient than silicon. This is "opportunity", they are already in use and will help create 5G and 6G phones.
"Compound semiconductors combine two or more elements of the periodic table, such as gallium and nitrogen, to form gallium nitride," Doran said. He explained that these materials are superior in terms of speed, delay, light detection, and emission. For silicon, this will help enable applications such as 5G and autonomous vehicles.
Although they may be used with ordinary silicon chips, compound semiconductors will enter 5G and 6G mobile phones, essentially making them fast enough and small enough while also having good battery life.
Doran said: "The advent of compound semiconductors has changed the rules of the game, and it has the potential to bring about change, just like the Internet is transforming the field of communications." This is because compound semiconductors may be 100 times faster than silicon, so they can grow for the Internet of Things The resulting surge in devices provides power.
What is quantum computing?
When you can have the superposition and entanglement of the quantum world, who needs the switching state of a classical computer system? IBM, Google, Intel, and others are racing to use qubits (also known as "qubits") to make quantum computers with powerful processing capabilities that far exceed silicon transistors.
The problem is that before realizing the potential of quantum computing, quantum physicists and computer architects have to make many breakthroughs. There is a simple test. Some people in the quantum computing community believe that before the advent of quantum computers, they need to meet their requirements: "Quantum first."
"It just means that on the road to Moore's Law, quantum machines are better at performing specific tasks than traditional semiconductor processors," Hampel said. So far, achieving that goal remains elusive.
What is Intel doing?
Since Intel is a pioneer in the manufacture of silicon transistors, it is not surprising that Intel has invested heavily in silicon-based quantum computing research.
"In addition to investing in expanding superconducting qubits that need to be stored at extremely low temperatures, Intel is also working on an alternative approach. The alternative architecture is based on 'spin quantum' Bit ', running in silicon. "
Spin qubits use microwave pulses to control the spin of a single electron on a silicon-based device. Intel recently used spin qubits on its latest "world's smallest quantum chip." Crucially, it uses silicon and existing commercial manufacturing methods.
Criddle explained: "Spin qubits can overcome some of the challenges presented by superconducting methods because they are smaller in physical size, easier to shrink, and work at higher temperatures. More importantly, spin qubits The design of the bit processor is similar to traditional silicon transistor technology. "
However, Intel's spin qubit system is still only close to absolute zero; cold computing will be closely related to the development of quantum computers. At the same time, IBM has a 50-bit processor Q, and Google Quantum AI Labs has a 72-bit Bristlecone processor.
What about graphene and carbon nanotubes?
These so-called magic materials may one day replace silicon. "Their existing electrical, mechanical, and thermal characteristics are far beyond what silicon-based devices can achieve," Doran said. However, he warned that it could take many years to prepare for the golden age.
"Silicon-based devices have been improved for decades and have evolved with the development of related manufacturing technologies. Graphene and carbon nanotubes are still at the beginning of this journey, and if they are to replace silicon in the future, they will The manufacturing tools needed for a goal still need to be developed. "
Atomic age
Whatever the future of other materials, we are now in the atomic age. "Everyone is thinking about atoms. Our progress has now reached the stage of counting single atoms, and even storage is looking for ways to work at the atomic level-IBM has shown possible ways to store data on a single atom," Harold said. "Today, creating a 1 or 0, a binary number used to store data, requires 100,000 atoms.
However, there is a problem here. Harold added: "As a means of storing or transmitting information, atoms are inherently less stable, which means that more logic is needed to correct errors." Therefore, future computer systems are likely to be the superposition of various technologies, each Each technique is designed to make up for the disadvantages of the other technique.
Therefore, no answer can extend the life of silicon to the next computing era. Compound semiconductors, quantum computing, and cold computing are all likely to play important roles in research and development. The future of computers is likely to be a hierarchy of machines, but so far no one knows what tomorrow's computers will look like.
"Although Moore's Law is coming to an end, the long-term and long-term trend of exponential computing power is unlikely to end," Hampel said.