EDA layout hot "cold" thinking

As one of the bottleneck technologies, EDA has received unprecedented attention. With the influx of multiple parties, the number of EDA companies in China has increased rapidly, and the number of financing cases has also increased.

First, the number of EDA companies continues to rise. According to statistics, there were 64 domestic EDA companies by the end of June. Secondly, the number of financings has increased with the increase in capital. In 2020, 9 companies completed 15 financings; in the first half of 2021, 6 companies have completed 8 financings. Finally, the scale of financing is larger. Since 2018, the financing volume of domestic EDA companies has increased rapidly. In addition, the IPO process of companies is also accelerating, including HuaDa JiuTian, ​​Prologix Electronics, and Guangli Microelectronics have all passed the review.

But beneath the apparent prosperity, new concerns are gradually emerging. Behind the "hot" boom, perhaps we need to think more calmly.

The mud and sand

As the cornerstone of the semiconductor industry, EDA is a typical technology, capital, and talent-intensive industry. Due to the long R&D and production cycle, the industry's talent demand and capital have become key factors in the development of the industry, which also reflects the main characteristics of the EDA industry. The three EDA giants Synaptics, Cadence, and Mentor (acquired by Siemens) have undergone more than 200 mergers and acquisitions in the past 30 years, forming an oligopolistic position.

Liu Weiping, chairman of Beijing HuaDa Emphasis Technology Co., Ltd., once mentioned in a speech that 7 of the top 10 EDA companies in the world are American companies, and HuaDa Emphasis barely made it into the top 10, showing the severity of my country's EDA industry being restricted by others.

Despite the increasing attention paid to EDA in the industry and the growing demand for domestic substitution in the context of great power competition, domestic EDA companies are making plans in the manufacturing, analog and digital fields. However, as the saying goes, there are two sides to a coin, and the emergence of too many companies has also caused the problem of dispersed power, and inevitably, the mud and the sand are mixed together.

Dr. Chen Chunzhang, a researcher at Pengcheng Laboratory and part-time professor at University of Chinese Academy of Sciences and Zhejiang University, pointed out that the domestic EDA industry is too hot and some chaos deserves attention: some manufacturers hang the EDA sign, but their business may not have much to do with EDA, or they just use EDA for chip testing, etc.; some companies have obtained EDA authorization, but they belong to the category of EDA technical services; and some companies use the EDA banner just to make it easier to obtain subsidies or apply for various types of funding.

From the perspective of specific positioning, as Chen Chunzhang pointed out, the front end of digital chips is becoming more and more complex, and the back end involves physical implementation, which needs to be combined with advanced processes. In this regard, EDA tools are the most complex, expensive, and difficult to make. In analog and RF design and implementation, engineers are required to understand circuit design and simulation, as well as layout skills and verification. In terms of system integration, EDA tools are required to plan an overall "electronic system design" architecture solution from system solutions, digital and mixed signal solutions, and advanced packaging and testing solutions in order to lead the development and catch-up of domestic EDA technology.

Therefore, we have to ask whether these newly established EDA companies are going to make point tools or system tools, whether they have technical team support, and whether they have in-depth understanding of the industry? Chen Chunzhang introduced that even point tools, that is, tools to solve a problem in the design process, are not simple, and they also require a systematic software, which requires the support of at least dozens or hundreds of people. Moreover, to do EDA, we need to communicate and discuss with companies in the circle to solve deeper problems; we need to establish an industry-university-research cooperation mechanism, share resources with scientific research institutes, and focus on talent training.

It should be noted that the industry's rush into the EDA industry is obviously related to the flow of funds and policy inclinations. But what makes people think deeply is that the industry has not paid attention to the important Matlab software, which is closely related to the EDA system simulation method. At present, students of related majors in major universities are basically using this software. If this software is banned, what should we do? Is there a solution in China?

As Liu Weiping said, EDA is actually a very difficult industry with high technical requirements. It also needs the cooperation of the industrial chain and the support of the ecosystem. It is not about solving the problem by doing EDA alone. Moreover, EDA is not something that any single company can solve alone.

Change is everywhere

As the main driving force of the EDA industry, Moore's Law has lost its momentum. With the advent of the post-Moore era, the EDA industry is also seeking innovation and change.

At present, most EDAs used are based on the foundation that began to take shape around 2000, which can be called "EDA1.0". However, with the advent of the post-Moore era, emerging technology forms such as 3D IC, Chiplets, SiP, new materials, and new devices require EDA to improve design efficiency and achieve methodological innovation in key links such as simulation and verification. At the same time, applications such as 5G+AIoT and smart cars are driving chip design towards online, digital, and intelligent, bringing about design trends such as cloud EDA, AI design, and software-hardware collaboration. Traditional EDA1.X methodologies and processes are difficult to adapt to the efficiency needs of chip design development, and EDA has also entered the era of comprehensive innovation in EDA2.0.

EDA 2.0 means that it will no longer be a combination of tools, but will integrate trends such as EDA+IP, EDA+AI, EDA+cloud, and move towards openness, intelligence, and cloud computing, becoming a complete service-oriented and customizable platform.

Wang Libin, Chairman and CEO of Xinhuazhang Technology, believes that the intelligent EDA 2.0 era will make chip design as simple as developing programs and chip manufacturing as flexible as building blocks.

Although my country's EDA industry faces challenges such as mainly providing point tools, shortcomings in advanced processes and packaging and testing, and a lack of continuously verified and iterated IP libraries, opportunities are also hidden in the challenges.

Some experts analyzed that the development of domestic EDA can focus on seizing the innovative opportunities of EDA tools spawned by new application fields, including the advancement of advanced processes such as 5nm and 3nm, new demands for compound semiconductors, new demands for chiplet design methods, and support for new applications such as optoelectronic integration. At the same time, domestic EDA companies with strong capital strength and technical reserves can make every effort to overcome the "hard bones" links such as front-end synthesis and back-end design under the collaborative cooperation of the industrial ecosystem. In addition, breakthroughs should be made in EDA sub-fields to create the strongest point tools and seek breakthroughs in EDA simulation and verification tools.

Rational Insight

Although the trend of EDA is clear, when you are in it, you also need to know what to do and what not to do.

Facing the trend of EDA on the cloud, Chen Chunzhang expressed his own views: Although cloud computing can solve some problems and improve the efficiency of simulation of some front-end systems, it involves security issues, and customers may not be willing to put data on the cloud, especially physically implemented data. I still remember that 20 years ago, customers proposed to share design and EDA tools through servers and networks, but in actual applications, there has not been much progress until today. In addition, the efficiency of putting some data on the cloud is not high, and it must be debugged on site by experienced engineers.

Industry insiders revealed that Foxconn had previously established a Design House and initially spent money to rent cloud EDA, but later gave up.

Therefore, the head of a well-known domestic EDA company also said that although moving to the cloud is a long-term trend, just like SaaS, the process is not simple, because EDA does not exist independently and requires the coordination of design and manufacturing processes, which is far from an overnight task.

Regarding the AI ​​level that is sought after by the industry, Chen Chunzhang also analyzed that the most effective machine learning is image processing and recognition. Currently, some products using machine learning methods are mostly used in layout and wiring optimization in digital circuits. Even though the research on algorithm concepts has made certain progress in academia in the past five years, it will take some time for them to be put into practical use in the industry. 

Regarding the EDA+IP combination, the above-mentioned person in charge pointed out that this combination has existed for many years. After the tools have been fully platformized, it is natural to develop IP because it can be sold in bundles, but this also requires stronger synergy to solve more industry challenges.

Chen Chunzhang also made an in-depth analysis of this. The digital circuit flow of EDA tools is not much different, but with the continuous improvement of integration brought by heterogeneous computing, the most complicated thing is how to use EDA to highly integrate different IPs, because IP protocols are becoming more and more complex. Whether it is USB, DDR, PCIe, etc., it involves a complete understanding of the system modeling of IP, chip design, and then to the application. IP design is not a new thing. Compared with SoC design or ASIC design, it ultimately belongs to IC design. EDA+IP can be bundled for business, so that EDA tools can be sold and IP solutions can be provided to customers. But fundamentally speaking, EDA plus IP cannot "create" myths. The essence is to improve the practicality of EDA tools.