Joinwin Electronics HK Limited518031About Join WinJoinwin Electronics HK Limited FLAT/RM 29, 22/F, YAN’S TOWER, 25-27 WONG CHUK HANG ROAD,ABERDEEN HONGKONG Shenzhen ZhongYiYingTong Technology Limited Room 1610, Block A, Overseas Decorative Building, Zhenhua Road,Huaqiang North Street, Shenzhen, 518031Join-Win will be your one-stop purchasing assistant. Join together, achieve win-win!Joinwin Electronics HK Limited518031About Join WinJoinwin Electronics HK Limited FLAT/RM 29, 22/F, YAN’S TOWER, 25-27 WONG CHUK HANG ROAD,ABERDEEN HONGKONG Shenzhen ZhongYiYingTong Technology Limited Room 1610, Block A, Overseas Decorative Building, Zhenhua Road,Huaqiang North Street, Shenzhen, 518031Join-Win will be your one-stop purchasing assistant. Join together, achieve win-win!Joinwin Electronics HK LimitedJoinwin Electronics HK LimitedJoinwin Electronics HK LimitedJoinwin Electronics HK LimitedJoinwin Electronics HK LimitedJoinwin Electronics HK LimitedJoinwin Electronics HK LimitedJoinwin Electronics HK LimitedJoinwin Electronics HK LimitedJoinwin Electronics HK LimitedJoinwin Electronics HK LimitedJoinwin Electronics

Home
News
Some Thoughts On AI's Impact On Chinese Integrated Circuits

Some Thoughts On AI's Impact On Chinese Integrated Circuits

April 02 , 2024 0

Ever since Robert Noyce first proposed the concept of integrated circuits in the 1950s, the industry has gone through a cycle of consolidation and fragmentation. In this process, many enterprises gradually grew through mergers and acquisitions, and eventually formed the current integrated circuit industry pattern.
At the "2024 China IC Leadership Summit" held at the same time of IIC Shanghai 2024 International IC Exhibition and Conference, Shi Longxing, professor of the School of Integrated Circuits of Southeast University and team leader of China IC Design Innovation Alliance, made a comprehensive analysis of the deep-seated characteristics of integrated circuits from three dimensions of industry, technology and talent. Help IC practitioners more accurately grasp its uniqueness and development trends.

Integrated Circuit Feature Review

First of all, from the perspective of industrial characteristics, the integrated circuit industry is undoubtedly a magnificent history of entrepreneurship. As an important player in the world IC industry, with more than 3,000 related enterprises, China may experience a more intense round of integration in the future to give birth to more competitive leading enterprises.

Another notable feature of the IC industry is product orientation. Integrated circuits are like a fruit tree, the fruit is a variety of products, and equipment, materials and design are the foundation to support the fruit tree, manufacturing is the backbone connecting the root and the fruit. The continuous strong market demand has promoted the integrated circuit industry to develop to a higher level, and it also requires the industry to make continuous progress in technology and innovation.

In addition, the value of integrated circuits lies in their application in the system. Whether it is a traditional PC, or today's popular mobile phones, or a variety of AI applications that may appear in the future, they cannot be separated from the support of integrated circuits. The changes and upgrades of system requirements directly promote the innovation and development of the integrated circuit industry.

However, the development of the integrated circuit industry has not been smooth sailing. It is deeply influenced by Moore's Law, which states that technological progress leads to cost reduction. This makes the IC industry a capital-intensive industry, requiring a large amount of capital investment to support research and development and production. At the same time, the trend of globalization has also brought opportunities and challenges to the integrated circuit industry. Although globalization contributes to the optimal allocation of resources and the rapid spread of technology, the uncertainties and challenges faced by globalization have also put forward higher requirements for the integrated circuit industry.

In terms of technical characteristics, the development of integrated circuits can be said to be a series of innovation processes. From the initial transistor to the current silicon plane process, each step embodies the wisdom and sweat of scientific researchers. In the future, the development of integrated circuit technology will show the trend of process size reduction, chip size increase, high energy efficiency and agility. These trends will push integrated circuits to achieve new breakthroughs in performance, efficiency and reliability.

At the same time, computing power is the basic demand for future technological development, which will put forward higher requirements for the integrated circuit industry. In order to meet the demand for high computing power for applications such as 6G and AI, the integrated circuit industry needs more exploration and practice in material and architecture innovation.

Finally, from the perspective of talent characteristics, the field of integrated circuits has extremely high requirements for talents. This is a multidisciplinary field, involving physics, chemistry, materials science, electronic engineering and other disciplines of knowledge. Therefore, the IC industry needs talents with interdisciplinary knowledge and innovative ability to support its continuous development. At the same time, with the continuous upgrading and change of the industry, the demand for talents in the integrated circuit industry has also shown a trend of diversification and compound. In the future, talents with practical experience, innovative spirit and interdisciplinary knowledge will become an important support for the development of integrated circuit industry.
Integrated circuit in the three dimensions of industry, technology and talent have shown unique characteristics, these characteristics together shape the unique face of the integrated circuit industry, and provide a broad space and unlimited possibilities for its future development.

The Impact OF AI On IC

With the rapid development of AI technology today, the enabling role of AI has penetrated into all walks of life. Many integrated circuit design companies have also adopted AI technology to improve design efficiency and performance. At present, there are a number of companies around the world focused on the development of AI chips, while other companies are widely involved in the field of AI.

The biggest impact of AI on chip design is reflected in the exponential growth of computing power, which far exceeds the performance gains that Moore's Law can bring. While Moore's Law has contributed to the growth of computing power, its gains in energy efficiency have been relatively limited, resulting in a "scissors gap" between demand and supply.

In addition to companies like NVIDIA developing general purpose AI, academia is also exploring new computing architectures to enable higher computing power. Analog computing, once superseded by digital computing, is now getting renewed attention in the age of AI. In recent years, analog computing has become a hot topic in ISSCC, IEDM, ISCA and other conferences.

The advantage of analog computing is that it requires less data handling, which is particularly prominent in digital copying computing. Analog computing improves energy efficiency by reducing data handling and using current summing to calculate values.

However, AI chips for analog computing are not yet mature and face several challenges:

Analog computing is difficult to meet the needs of diverse networks, because AI network design tends to be complicated, while analog computing is more regular.

It is difficult to strike a balance between precision, energy efficiency and area efficiency in simulation calculation, which usually requires the combination of multiple domains to solve this problem.

Analog computing inevitably involves analog-to-digital conversion, and the loss of conversion circuit and peripheral reference circuit becomes a new bottleneck.

Professor Shi Longxing said that his team has carried out research for many years to explore the simulation computing architecture and algorithm of multi-operator fusion, as well as the design of the entire simulation chip and demonstration system.

He said: "In terms of storage and computing, we know that the von Neumann architecture of digital computing is characterized by its powerful storage and power consumption. In order to reduce power consumption, we consider memory computing and process computing. We compared different technology routes and finally chose to use SRAM for in-memory computing."

In addition, Professor Shi Longxing's team also studied CRM, AI accelerator, compiler design and COC chip, and achieved remarkable results.

Professor Shi Longxing also stressed that the impact of AI on integrated circuits is not only reflected in the huge demand for computing power, but also in the manufacturing process.

Taking the research of Academician Wu Hanming of Zhejiang University as an example, virtual manufacturing significantly reduces the dependence of integrated circuit research and development on the design of the flow sheet, shorting the time and reducing the cost.

It is worth mentioning. Last year, Nature published an article on human-machine collaboration to optimize semiconductor process research and development, through human-machine interaction, combined with human experience and machine learning, process development can be more efficient, and this approach has shown advantages in cost and benefit.

AI Helps The Next Generation OF EDA Break The Development Bottleneck

The evolution direction of the next generation of EDA will undoubtedly focus on intelligence and agility, and the integration of artificial intelligence technology may become a key weapon to crack the bottleneck of EDA development.

According to Professor Shi Longxing, at present, the academic community is starting from the whole process of integrated circuits and exploring the possibility of how AI can empower EDA from top to bottom. "Of course, we also clearly recognize that not all aspects are suitable for AI intervention, but in general, AI has played a positive role in multiple levels of EDA," he said.

In terms of progress.

First, in the area of High-level synthesis, machine learning techniques are particularly widely used. This is mainly reflected in two aspects:

One is the prediction of the design result,

The second is to assist in the exploration of the design.

Among them, machine learning has shown great potential in the evaluation and analysis of results.

In addition, Cross-Platform learning is also providing strong support for EDA tools.

In the design process, active learning is being used to improve the iterative algorithm process, assist in selecting the appropriate synthesis algorithm, and classify the synthesis process.

In the field of physical design, the application of machine learning is equally rich. It mainly plays a role in the layout (Placement), Routing (Routing) and power network design, especially in the prediction of routing information and IR Drop prediction, artificial intelligence has shown its indispensable value. In addition, the design of 3D interface layer is also a research hotspot in EDA field, and many remarkable results have been achieved.

In manufacturing, processes including lithographic hotspot detection OPC and resolution assisted graphics insertion have also begun to introduce AI technology.

In the field of testing, AI technology is mainly used to solve the problem of test-level redundancy and reduce test complexity.

The optimal design of analog circuits is also the focus of AI technology. This is mainly reflected in the optimization before the board drawing, the auxiliary board drawing generation and the simulation after the board drawing generation. Through the prediction of birth control parameters and the optimization of circuit architecture, AI technology is bringing revolutionary changes to simulation design.

In the industry, many well-known companies such as Synopsys, Cadence, and Siemens have fully introduced AI technology in their tools. In addition, Ansys also incorporates AI technology in multi-physics simulations. It is worth mentioning that the United States has already begun to lay out the application of AI in the EDA field in research institutions such as MSF SRC and DARPA.

In summary, artificial intelligence is injecting new vitality into the future development of integrated circuits.