As the foundation of the intelligent society and the engine of the digital economy, the importance of computing power has been raised to an unprecedented height. At present, the improvement of computing power still faces challenges from multiple dimensions. Whether it is at the hardware level or the architecture level, the development of computing power urgently needs changes.
Advanced computing technology refers to all innovative technologies that effectively increase the scale of existing computing power, reduce computing power costs, and improve computing power utilization efficiency from the aspects of computing theory, computing architecture, and computing system.
In order to solve the computing power challenge and break the computing power crisis in the post-Moore era, advanced computing technology will make breakthroughs from two directions: the improvement of single-point computing performance and the efficient use of computing power systems.
Through scientific research data analysis, expert interviews and other methods to gain insight into the development trend of the advanced computing field, explore advanced computing technologies with disruptive technologies and commercial prospects, and extract seven major trends in advanced computing technologies and applications in 2022 for readers.
Computing power is the carrier of multi-field and multi-technology integration, and the pursuit of computing power is endless. In the future, with the continuous development of advanced computing technology, computing efficiency and energy efficiency will usher in new breakthroughs, and the computing power system will undergo subversive changes.
With the support of advanced computing technology, more intelligent and intelligent applications will emerge, and green, low-carbon, open and open source computing power will also become the trend of ecological co-construction in the digital economy era.
At present , quantum computing has entered the NISQ era. As the core of quantum computers, quantum chips ushered in a big explosion in 2021. According to statistics, quantum computing hardware and software companies completed a total of 42 financings in 2021, with a total financing of about 2.691 billion US dollars, exceeding the previous total.
At the end of 2021, IBM launched the world’s first 127 superconducting qubit chip “Eagle”, successfully breaking through the 100-bit qubit mark of superconducting quantum chips.
To enter the era of general quantum computing, the quantity and quality of qubits, the connectivity between bits, and fault tolerance still need to be further breakthroughs. The international academic community generally believes that it will take 15 years or more to truly realize a programmable universal quantum computer, but with the promotion of policies, the upgrading of related industries and the continuous support of capital, large-scale and commercial quantum computing will accelerate.
Silicon Photonics Chip
At Moore’s end, “optical”
silicon photonic chips have the technical advantages of both microelectronics and nanophotonics. They not only have the characteristics of small size, low cost and high integration of microelectronics, but also combine the advantages of high bandwidth, high speed and multi-channel optoelectronics. It is fully exerted, and the efficient integration of the two physical systems is realized.
At present, silicon photonics chips are still faced with the constraints of the industrial chain and process level, and a systematic design method and standardized manufacturing process have not yet been formed, and the density of optical devices needs to be further improved.
The commercialization of silicon photonics chips is an ongoing and lengthy process. In the next three years, commercial products of silicon photonics chips will first appear in application scenarios with specific requirements for computing power, delay and power consumption, such as quantitative transactions, autonomous driving and big data centers. In the next 10 years, with the continuous evolution of technology, optical computing will be popularized on a large scale based on the long-term complementary coexistence of silicon photonic chips and electronic chips.
Answers from the brain The
human brain is the most perfect information processing system in nature. Brain-like computing draws on the way the human brain processes information, subverting the traditional computing paradigm with fewer devices, lower energy consumption and faster speed. The brain-like chip is the hardware carrier for the application of brain-like computing, and has the characteristics of “event-driven” and “integration of storage and computing”.
At present, the brain-like chips that can be mass-produced basically still use circuits to simulate artificial neural synapses to realize brain-like computing, which not only requires high manufacturing process, but also has very low production efficiency. Exploring more ideal materials will be the key to breakthroughs in brain-like chips.
In the future, the development of artificial intelligence will inevitably move from narrow and weak artificial intelligence to more robust and versatile artificial general intelligence. The deep integration of brain-like chips and silicon-based computing architecture will provide the necessary conditions for the realization of artificial general intelligence.
With the popularization of cloud computing, data centers are facing problems from performance, scale, energy consumption and security while expanding their capacity rapidly. In order to better solve the bottleneck of the data center and realize the cost reduction and efficiency increase of cloud services, DPU (data processing chip) has become the perfect solution to replace the traditional smart network card. At present, the core competitive element of DPU chips is the ultimate balance of performance and programmability.
The data center is moving from the “CPU-centric” era to the “data-centric” era. Data is a main thread that runs through the three elements of computing, network and storage. In the future, DPU, CPU, and GPU will become the troika in large data centers. As the brain of the entire data center, DPU links and manages the entire computing system through data streams.
The key hub of cloud-edge-device collaboration
Distributed cloud is close to the source of data, mounted on the edge infrastructure between the data center and terminal devices, and can provide low-latency, wide-connection, and high-bandwidth cloud services. At present, distributed cloud still faces difficulties in deployment cost, heterogeneous management of software and hardware, and security system.
With the development of hardware devices, distributed cloud nodes will have the capabilities of part of the central cloud, but they will not completely replace the central cloud, but build a more complete cloud computing architecture in the form of cloud-side-end deep collaboration. In the future, distributed cloud will serve applications that have stringent requirements on latency and connectivity, such as autonomous driving, industrial manufacturing, and smart cities, unlocking more value of edge intelligence.
Ubiquitous computing power
Ubiquitous computing power network
With the increasingly prominent contradiction between the explosive growth of data and the performance limit of a single point of computing power, it has become a general trend to connect ubiquitous computing power with the network and build a three-dimensional computing power network that integrates cloud, edge and terminal.
In the future, computing power will show a trend of ubiquitous distribution, multivariate heterogeneity, and location insensitivity. With the parallel development of network technology and computing technology, and the deep integration of computing networks, the arrival of an intelligent and schedulable ubiquitous computing power network will accelerate.
In the next 10 years, computing power will penetrate into every aspect of human production and life like basic resources such as water, electricity, and coal, and become the core element of the operation of an intelligent society.
Advanced computing accelerates scientific renaissance
Every innovation in computing power boosts the development of science and technology. Entering the era of computing power, with the continuous improvement of global computing power, the momentum of the application of computing technology to scientific research is also increasing day by day. As the cornerstone of scientific research, computing technology has solved many scientific problems such as protein folding for human beings, and has also opened up a new field of research.
We believe that in the future, advanced computing technologies represented by quantum computing and optical computing will be equipped with new engines for artificial intelligence and other tools to accelerate the renaissance of human science.