Black Sesame Huashan No. 2 chip released, becoming the only domestic chip that can support L3 autonomous driving

Intelligent driving is the core productivity of human society in the future.

According to the “Internet of Vehicles (Intelligent Connected Vehicles) Industry Development Action Plan”, “Automotive Industry Medium and Long-term Development Plan” and other policies, by 2020, the loading rate of new car driver assistance systems (L1) will reach more than 30%; The DA (Assisted Driving) and CA (Conditional Autonomous Driving) new car assembly rate will reach 80%. According to the current price of 50-150 US dollars for low-level intelligent driving chips, the market size of intelligent driving special chips can reach 10 billion US dollars in the short term. In the long run, this figure will rise to the scale of hundreds of billions of dollars.

At the just-concluded 2020 Beijing International Automobile Exhibition, BlackSesame, which made its debut at the China Auto Show, not only brought its self-developed vehicle-grade chips, Huashan No. 1 A500, Huashan No. 2 A1000, and fusion perception high-precision positioning products , as well as the latest products such as the intelligent cockpit driver monitoring system jointly built with Bosch, and also released the FAD (FullAutonomousDriving) fully automatic driving computing platform based on the A1000 chip for the global market, which has become the most noteworthy in this auto show. One of the innovative forces of intelligent driving.

Founded in 2016, Black Sesame Intelligent Technology is an enterprise focusing on visual perception technology and independent IP chip development. The main areas of focus are embedded image and computer vision, and the core business is to provide a computing platform based on image processing, computing images and artificial intelligence perception chips to provide complete commercial solutions for ADAS and autonomous driving.

In April 2019, Black Sesame completed the B round of financing. The main investment institutions include Northern Lights Venture Capital, Junhai Chuangxin, SAIC Group, SK China, China Merchants Capital, Core Dynamics, etc. In August of the same year, Black Sesame’s first vehicle-grade intelligent driving chip, Huashan No. 1 A500, was launched in China and started mass production; in June 2020, the second car-grade intelligent driving perception chip, Huashan No. 2 A1000, was released and became the only one Domestic chips that can support L3 autonomous driving are expected to start mass production at the end of 2021 at the earliest.

Black Sesame Huashan No. 2 chip released, becoming the only domestic chip that can support L3 autonomous driving

On the eve of this auto show, Black Sesame also delivered the first batch of fusion perception high-precision positioning products based on the A500 chip to SAIC, mainly for the decimeter-level positioning needs of L2-L3 autonomous driving. In addition, Black Sesame and SAIC, BYD and other domestic leading car companies are also in full swing for L2+ and L3 level autonomous driving projects.

The traditional automobile industry chain begins to be restructured

In February 2020, 11 national ministries and commissions jointly released China’s national strategy for smart cars, which clearly identified 2025 as an important time node for the formation of China’s independent industrial chain and technical standards. Breakthroughs include complex system architecture, complex environment perception, and intelligent decision-making control. , human-machine interaction and human-machine co-driving, vehicle-road interaction, and network security, including a number of forward-looking key technologies, to promote large-scale and comprehensive application pilots of city-level smart vehicles in qualified regions, and to support advantageous regions to create a national Internet of Vehicles Pilot area.

In an interview with “Electronic Engineering Special”, Liu Weihong, COO of Black Sesame Intelligent Technology, said that vehicle-mounted high-precision sensors (camera, lidar, millimeter-wave radar), vehicle-level chips, intelligent operating systems, vehicle-mounted intelligent terminals, and intelligent computing platforms are in The intelligent driving industry in China is a new trillion-level industrial innovation opportunity for China. It can not only drive the entire automobile industry out of the traditional production and manufacturing model, but also step into the core chip, artificial intelligence, battery technology, big data, operation The era of intelligent driving/autonomous driving represented by management has also brought changes to the traditional industrial chain while promoting the upgrading of the automobile industry.

Raw material suppliers, second-tier suppliers, first-tier suppliers, OEMs, dealers, and finally sales outlets. This is the ecological chain model of the traditional automobile industry, which has not changed for many years. However, in Liu Weihong’s view, as more and more OEMs begin to cultivate their own parts suppliers and increasingly focus on “travel solutions”, the transformation of the traditional ecological model has begun.

“The gradual establishment of the autonomous driving ecosystem has a great impact on the traditional first-tier suppliers. Many vehicle companies now prefer to communicate directly with second-tier suppliers and find first-tier suppliers for production after forming solutions. At the same time, travel solution companies The massive outflow of the black sesame has also put a lot of pressure on the depot.” Therefore, Shan Jizhang, CEO of Black Sesame Intelligent Technology, believes that the value of Black Sesame’s existence is that by combining algorithms, operating platforms and chips, and cooperating with first-tier suppliers, After forming a complete solution, it will serve the OEM.

However, since the development capabilities of different first-tier suppliers and even OEMs vary greatly, the solutions provided by second-tier suppliers like Black Sesame must have sufficient “flexibility”. That is to say, for upstream manufacturers with strong development capabilities, Black Sesame can provide them with separate chips, tool chains, software, and development environments, and the system integration work is done by the other party; for manufacturers with relatively weak capabilities, Black Sesame must provide Complete “turnkey” solution.

Create the “strongest brain” for intelligent driving

For the development of autonomous driving systems, the industry generally believes that the progressive route from L2+ assisted driving to L4/L5 autonomous driving is the most feasible path. This requires the corresponding computing platform to have super scalability, support the smooth evolution of system development, meet the differentiated requirements for computing power and power consumption of autonomous driving at all levels, and improve the development efficiency of partners such as OEMs.

The FAD fully automatic driving computing platform released this time is based on the cascade scheme of two Huashan No. 2 A1000 chips. The two-chip system can work completely independently, and supports the redundant design of power supply and video acquisition system. The computing power can reach 70TOPS- 140TOPS, fully meet the design requirements of front-loading products.

According to the single chapter, FAD is “the only intelligent driving platform that can benchmark Tesla’s FSD”, which can fully meet the computing needs of multi-level systems from L2+ assisted driving to L4/L5 autonomous driving.

According to the plan, Black Sesame will also provide a four-chip combination solution in the future, the computing power will reach 280TOPS, and the overall energy efficiency ratio will be as high as 6TOPS/W. With the self-developed vehicle-level operating system and multi-chip cascaded real-time software platform framework, FAD platform It will be able to meet the needs of intelligent driving solutions in different scenarios such as passenger cars, commercial vehicles, and work vehicles. At the same time, the FAD platform also supports multi-channel high-definition sensors and has rich interfaces (for example, a single A1000 chip can support 12 cameras at the same time), open rich API capabilities through FADSDK, and provide open automatic neural network optimization tools.

“The level of peak computing power is only a part of the car chip’s attention. This is like a ‘decathlon’ player, not just a single ability that is outstanding, but to achieve a comprehensive balance.” In addition to computing power, Dan Jizhang believes that car regulations The design difficulties of high-level chips also include information security, functional security, heterogeneous architecture design, processing of different data types, thermal management and other aspects. At the same time, considering that software-defined cars have become an industry consensus, it is also necessary to reserve enough redundant space in the design to cope with the continuous changes in car architecture and AI algorithms.

As the core main chip of FAD’s fully autonomous driving computing platform, the Huashan No. 2 A1000 adopts TSMC’s 16nm process technology and has completed tape-out in April this year. Compared with the predecessor Huashan No. 1 A500 chip using 28nm process, the computing power of A1000 has increased by nearly 8 times, reaching 40-70TOPS, the corresponding power consumption is 8W, and the energy efficiency ratio exceeds 6TOPS/W. The chip field is in a leading position.

The chip is based on the self-developed multi-layer heterogeneous TOA architecture of Black Sesame, and integrates image sensing technology with independent intellectual property rights, image and video compression coding technology, computer vision processing technology and deep learning technology IP. A1000 has 8 built-in CPU cores, including DSP digital signal processing and hardware accelerators, and supports the access of autonomous driving sensors including lidar, millimeter-wave radar, 4K camera, GPS, etc. Sensor fusion scheme. At the same time, in order to meet the requirements of vehicle-road collaboration and vehicle-cloud collaboration, A1000 also integrates PCIe high-speed interface and vehicle-grade Gigabit Ethernet interface, which is convenient for developers to carry out integrated development.

In terms of security, the Huashan No. 2 A1000 chip itself has been designed with an independent vehicle-level safety island. Through the R-Lock dual-redundant interlocking architecture and multiple reliability designs, it provides a comprehensive information security system, supports hardware encryption, and meets ASILB/ D and CCEAL5+ vehicle-level safety certification requirements.

Shan Ji Zhang said that the intelligent driving industry has gradually returned to rationality from the initial madness, and has begun to return to mass production from concepts, which is why more people in the industry have begun to talk about the L2+/L3- stage. He hopes that by 2025, Black Sesame can occupy 30% (about 3 million vehicles) of the L2+ to L3 autonomous driving market, and at the same time emphasized that “opening” will eventually replace “closed” and become the era of intelligent driving. Key words. At present, whether it is software-defined cars or opening chips, algorithms, and product iteration capabilities to customers, they are all following this principle without exception.

Responsible editor: tzh

The Links:   LMG9980ZWCC-01 MG75H2CL1

Related Posts