Tanjing Technology: Solve the "storage" problem of AI chips and become a full-stack terminal intelligent solution provider

【) Beijing】Reported on March 26 (Text/Du Baojie)

In the field of chip design, there is an IC company that is as powerful as Qualcomm, MediaTek, AMD and other giants, Marvell.

Marvell was founded in Silicon Valley in 1995. In 2000, this Chinese-founded chip company was listed on the Nasdaq. Later, it gradually became a top supplier of storage, communication and consumer semiconductor solutions, selling more than 1 billion every year. Marvell once became one of the top five semiconductor companies in the world. Currently, Marvell still occupies a pivotal industry position in the fields of storage, communications, smartphones, and consumer electronics semiconductor solutions. At the same time, Marvell has also "cultivated" many chips for the industry. Talent.

Lu Yong, founder and CEO of Tanjing Technology, previously served as a technical executive at Marvell. He was born in the chip field and devoted himself to the chip field. In 2017, the first year of the outbreak of artificial intelligence, domestic AI chips suddenly emerged and ushered in the best period of development. This is especially true for Lu Yong.

"I have worked at Marvell for ten years since 2007. Unlike many large companies that only do peripheral work in China, we bring a lot of core chip technology to China, so I also I have single-handedly led a relatively strong chip R&D team in China. The entrepreneurial environment in China has been so good in recent years, and I am in the midst of a big trend. I always feel that I should do something."

When choosing the direction of entrepreneurship, various fields of artificial intelligence have become a boom in entrepreneurship. Lu Yong said that the AI world is a science fiction dream he has had since he was a child. When he discovered that AI technology has become practical and usable, he was even more convinced that it would be his future 5. It's a career worth investing in for 10 years or even for a long time. Artificial intelligence will become ubiquitous. In Lu Yong’s view, a complete artificial intelligence scenario can only be formed when every device around him truly has artificial intelligence functions, instead of requiring remote support such as cloud services. Therefore, At the beginning of his business, Lu Yong was very sure of his technical route. To do artificial intelligence, he had to develop the terminal and make the industry's top terminal artificial intelligence chip.

It is understood that Tanjing Technology is better at producing native chip technology than its peers. Lu Yong explained: "Native technology means that we will not purchase or improve other companies' ready-made solutions. Most of our team's talents are from Tsinghua University. The core members basically have 10 to 20 years of experience in the chip industry. They will learn from the latest solutions." We start from the bottom, from the chip design, the structural framework of the chip, the system based on the chip after the chip is released, the software algorithm based on this system, etc., we will all independently develop and do it, and we will not do simple integration."

Lu Yong believes that unlike entrepreneurship in the Internet era, the threshold for making chips is actually much higher. Senior industry insiders generally agree that it requires years of industry accumulation and R&D experience. In particular, AI chips test the comprehensive capabilities of the team. Many outsourcing practices actually do not have real core competitiveness.

It is understood that the core module of Tanjing Technology’s AI chip has been designed. The next step is productization, converting the chip IP into a real chip. It is expected that there will be a specific chip mass production plan in the second half of this year. Officials revealed that the current chip IP indicators are very advanced compared with similar AI chips in the entire industry. Specific test data will be disclosed when it is officially released.

When making terminal AI chips, Tanjing Technology has its own focused application scenarios. Security and smart home are the market directions that the team first considers. Of course, in the middle and later stages, it will also try to expand its layout in the field of autonomous driving chips. Lu Yong said that focusing on the team's core technology, it is unrealistic for one chip to conquer the world. It is still necessary to make products based on customers' application scenarios and grasp their own R&D rhythm. In fact, there may be differences in terms of application fields, but as far as AI chips are concerned, In terms of the core technology, it has at least 50% commonality. The commonality of the core technology means that the time required for the team to switch from the first product to the second product is very fast, and there is no need for a complete overhaul. It is developed from scratch, and targeted optimization adjustments can be transferred to another scenario for use.

Lu Yong believes that in the early stages of a company, technology and business are the breakthrough point in facing the complex competitive environment and are the barriers to team competition. Making chips is a systematic project, which requires no shortcomings in every link. The Tanjing Technology team is characterized by being both balanced and experienced, from creativity to industrial experience, from software algorithms to hardware design. This is the core Advantage.

"Many people rush into the AI chip industry to only superficially solve the superficial problems, but this matter is not that simple. In fact, the difficulty of the entire AI chip industry is not to solve the problem of computing power, but to solve the problem of 'storage', how to better Manage this data, schedule this data, and then improve performance and reduce power consumption. This is the crux that most people have not solved. In addition, many people are doing embedded AI algorithms and making a lot of simplifications from the algorithm level, but this cannot be done. To the best level of industrialization and practicality, all information processing and calculations in future terminals will need to be run on AI chips, and companies that cannot solve these problems well are just playing games."

In addition to solving the "storage" problem of AI chips through a complete set of R&D and design, Tanjing Technology will also package the software algorithm solution as a whole and provide customers with full-stack services. Customers can use it right away and use it in smart security Embedding in products such as cameras or smart home appliances allows customers to develop more easily. In the future, Lu Yong hopes that the company will grow into an AI technology-based company, including all aspects related to AI technology. It will be a solution provider that provides all solutions from underlying chips to upper-layer algorithms and software for various customer demand scenarios.

"In the entire artificial intelligence trend, AI chips are a relatively low-level but very core important component in the industry. There is no AI without chips, but China's chip design level is still relatively low. On the one hand, our participation will help the domestic AI chip industry to be better If we continue to implement it, another point is to develop ourselves by taking advantage of the tide and complement each other with the industry for win-win results." Lu Yong concluded at the end of the interview.

Project: Terminal AI chip
Company: Beijing Tanjing Technology Co., Ltd.
Official website:www.intenginetech.cn