In July 2022, China’s new energy vehicles sold out. On July 22, the Passenger Vehicle Market Information Joint Conference (referred to as the “Passenger Federation”) released data showing that the retail sales of new energy vehicles in July are expected to reach 450,000 units, a year-on-year increase of 102.5%.
Tian Yulong, spokesman and chief engineer of the Ministry of Industry and Information Technology, said recently that since the beginning of this year, China’s new energy vehicles have achieved steady and rapid development, with two major highlights: First, the scale of production and sales reached a new high. In June, the production and sales of new energy vehicles completed 59% of the total. 10,000 units and 596,000 units, an increase of 1.3 times year-on-year. In the first half of the year, China’s new energy vehicle sales reached 59% of the global market, nearly double the level in Europe. Second, new breakthroughs have been made in technological innovation. The newly developed lidar, domestic chips, and vehicle-based basic computing platforms have all been applied in vehicles, and the level of technological innovation has been continuously improved.
In May, Black Sesame Smart, a domestic large-scale computing power vehicle-level chip manufacturer, reached a platform-level strategic cooperation with Jiangqi Group, and a number of Sihao brand mass-produced models will be equipped with Huashan No. 2 A1000 chip. On July 18, Horizon announced the deepening of strategic cooperation with SAIC Motor and Zero Beam Technology. Based on Horizon’s high-performance and large-computing AI chips, they will jointly build a future-oriented intelligent driving computing platform and promote the development and development of high-end intelligent driving products. It is expected that the mass-produced cooperative models equipped with Horizon Journey 5 will start to land in 2023.
In 2022, with the acceleration of automobile intelligence and electrification, the global automobile industry pattern is also changing, and Chinese companies have entered the main track one after another. What is the development stage of China’s new energy vehicles now? SAIC and GAC are testing higher-level autonomous driving. How can large computing power chips speed up the ride? What are the latest developments of Chinese manufacturers in the fields of lidar and automotive computing power chips? This paper analyzes in detail.
China’s new energy vehicle shipments account for 59% of the world’s total, detonating the future smart driving chip market
According to the statistics of CleanTechnica, from January to June 2022, the cumulative sales volume of new energy passenger vehicles in the world is 4.12 million, of which 2.51 million are sold in China, with a global market share of 59%, surpassing 820,000 in Europe and 820,000 in North America. 410,000 vehicles. In the TOP20 global new energy vehicle brand sales list from January to June 2022, there are 12 Chinese brands, more than half of the country; in the global new energy brand model sales list TOP20, Chinese brand models account for 70%.
The China Automobile Association predicts that sales of new energy vehicles are expected to reach 5.5 million in 2022, a year-on-year increase of more than 56%. According to research data from CICC, by 2025, the penetration rate of high-level autonomous driving will reach 65.5%, and the market demand for smart driving chips in China will reach 13.83 million pieces. In the next 3-5 years, intelligent driving will usher in a period of rapid development, and the automotive industry will usher in huge opportunities.
In the field of automotive computing power chips, foreign chip manufacturers such as Europe, America and Japan control key technologies. The self-sufficiency rate of domestic automotive computing power chips is less than 10%, and the localization rate is less than 5%. With the continuous advancement of automobile intelligentization, and under the background of low self-sufficiency rate and low domestic production rate of automotive-grade computing power chips, major chip manufacturers have deployed large-scale computing power chips for autonomous driving (the computing power of a single chip is greater than 100TOPS). The “Battle of Computing Power” in the automotive field in the chip industry has officially started.
Shan Jizhang, CEO of Black Sesame, told the media that in the field of smart electric vehicles, Chinese car companies lead the world in development speed, and the local supply chain system is the key to supporting the rapid technological iteration of Chinese self-owned brand car companies. Intelligent driving must rely on the support of large computing power chips, and only by mastering the core IP can rapid iteration of the chip be achieved.
The localization process of large computing power chips is accelerated, and 2023 will be the year of concentrated mass production
Cai Weijie, director of the software development department of Zhejiang Geely Automobile Research Institute, said that car users’ pursuit of a better life hopes that cars can be personalized and become the third space for activities. With the development of car networking and intelligence, people hope that cars can share intelligent life. Future cars will have three new attributes: 1. Rapid iteration, users expect cars to be often new; 2. Ecological integration, with the help of the current Internet of Things and 5G technology, to realize the integration of everything; third, data intelligence, with the help of intelligent perception, big calculation Power chip and AI technology to promote the evolution of data intelligence.
With the continuous development of automobile intelligence, the computing power required by autonomous driving technology has shown an exponential explosive growth, and chips have become an important part of the development of the automobile industry. From L2, L3, L4 to L5, each step up will increase the computing power demand by at least 10 times. This means that the demand for automotive chips, especially large computing power chips, will maintain a rapid growth trend.
The industry generally believes that the computing power required by the car can be comprehensively calculated based on the amount of data collected by the sensors in the car. Among them, L2-level autonomous driving requires roughly 10TOPS of computing power; L3 requires more than 100TOPS of computing power; to L3+ computing power level Has risen to more than 1000TOPS. In the field of automotive-grade large-scale computing power chips, in addition to Nvidia, Qualcomm and other established brands, the market monopoly has been gradually broken. Brands represented by Huawei, Horizon, and Black Sesame have joined the battle. In 2022 or 2023, they will become China-made large-scale computing power chips. production year.
Recently, Wang Jun, President of Huawei’s Intelligent Vehicle Solutions Product Department, said that under the new industrial division of labor and cooperation model, Huawei adheres to the “platform + ecology” development strategy, and Huawei focuses on ICT technology, focusing on the three major platforms of iDVP, MDC and HarmonyOS smart cockpit. , to build an automotive ecosystem.
In 2021, Huawei’s cooperative passenger car project has been fully launched. Including the GAC AION LX Plus using Huawei’s MDC intelligent driving platform, and the Changan Group’s Great Wall Mecha Dragon, equipped with dual MDC chips. Huawei MDC (Mobile Data Center) is positioned as a computing platform for intelligent driving. Huawei MDC810 has reached 400TOPS computing power. Li Zhenya, general manager of Huawei’s smart car solution BU MDC field, revealed that in 2022, Huawei will also release the MDC100, an intelligent driving computing platform.
According to Horizon’s official WeChat account, Journey 5 is the third-generation vehicle-grade product launched by Horizon for high-level autonomous driving application scenarios, featuring both high performance and large computing power. Journey 5 has a single-chip AI computing power of up to 128TOPS, and is equipped with Horizon’s next-generation BPU Bayesian deep learning acceleration engine, giving full play to the technical advantages of software and hardware collaborative optimization to achieve industry-leading and sustainable growth in real AI computing performance. Journey 5 is also the most mature mass-produced 100-TOPS-level domestic AI chip with large computing power. It is reported that Horizon’s Journey 6 chip based on the new BPU Nash architecture is under development, and the AI computing power can be increased several times, which will fully meet the future higher-level computing requirements for the entire vehicle intelligent center, and help accelerate the intelligent transformation of the automotive industry.
The intelligent driving computing platform, which will be mass-produced by SAIC and Horizon in 2023, will be based on the Dual Journey 5 chip by Zero Beam Technology, which integrates full-stack electronic architecture, high real-time software computing platform, industry-leading BEV algorithm, and global closed-loop data factory. Created with technology, it is the core component of the technical base of SAIC’s “Galaxy” solution, covering high-speed, parking and urban full-scene experience, and finally realizing data-driven, end-to-end integrated intelligent driving function products.
Shan Jizhang, CEO of Black Sesame Intelligence, said that Black Sesame Intelligence is committed to building core chips for autonomous driving, and has built core competitive advantages based on two independent and controllable core IPs, including vehicle-level image processing ISP and vehicle-level deep neural network accelerator NPU. , so that the vehicle “sees clearly” and “understands”.
The Huashan No. 2 A1000 Pro released by Black Sesame is a car-grade autonomous driving computing chip with leading domestic performance and computing power. It adopts heterogeneous multi-core architecture, 16-core Armv8 CPU, 16nm process, and can support L3/L4 high-level autonomous driving functions. , has been successfully taped out. A1000 Pro is built on the basis of two self-developed core IPs of Black Sesame Intelligence – automotive-grade image processor NeuralIQ ISP and DynamAI NN automotive-grade low-power neural network acceleration engine. Thanks to the large computing power architecture of DynamAI NN, A1000 Pro supports INT8 is sparsely accelerated, the computing power of INT8 is 106TOPS, and the maximum computing power of INT4 can reach 196TOPS, leading in the highest computing power of domestic autonomous driving chips.
L3 level autonomous driving advances, lidar accelerates
Lidar has the characteristics of long distance measurement, excellent angular resolution, and little influence by ambient light, and can directly obtain the distance and orientation information of objects without the need for deep learning algorithms. Compared with other sensors, these advantages can significantly improve the reliability of the automatic driving system, so they are considered by most OEMs and Tier1 suppliers to be a must for L3 and above automatic driving (the responsible party is the vehicle system when the function is turned on). sensor.
“There are some commercial vehicle companies that are firmly moving towards L2 autonomous driving. For example, Pony.ai Robotaxi is pushing, and Weyuan has also cooperated with Bosch. Everyone thinks that the commercial space of L2 level is limited, and the realization must be expanded to L4 level; Mercedes-Benz The Mercedes-Benz L3-level conditional autonomous driving system took the lead in passing the strict technical regulations approval of the German Federal Motor Transport Authority in 2021, becoming the first automobile company in the world to obtain the UN-R157 certification of the United Nations regulation. This year, we learned that there are 8 domestic Home OEMs are using lidar, all of which are preparations for L3-level autonomous driving or higher-level autonomous driving, and the trend is irreversible.” Duan Bobo, co-founder and CTO of Yingchi Technology, summed up one of the trends in automotive transformation.
2022 will be the first year for the popularization of L3, and the biggest change it will bring is the large-scale installation of mass-produced models of lidar products. Take the Avita E11, the first model jointly built by Changan, Huawei and CATL as an example. This car is equipped with CATL’s latest-generation CTP technology ternary lithium battery pack, and is equipped with Huawei Drive ONE electric drive system. It takes 3.98 seconds, and is equipped with 3 lidars, 6 millimeter-wave radars, 12 ultrasonic radars, and 400TOPS computing power supplied by Huawei, which can achieve higher-level assisted driving capabilities. The NIO ET7, which began to be delivered in March this year, is equipped with the world’s first mass-produced 1550nm lidar, and it is also the lidar with the farthest detection distance in mass production.
On July 13, lidar supplier RoboSense (RoboSense) stated on its official WeChat account that in the first half of this year, the company’s order volume increased by more than 10 times compared with the same period last year. The company’s product strategic planning director was interviewed by the media. It was revealed that at present, the company’s multiple intelligent production lines in Guangzhou, Shenzhen and other places have been deployed, and the annual production capacity will reach about 1 million units by the end of this year.
Lidar has almost become standard in many new domestic car-making forces and many new cars under its own brands. Taking Weilai ET7 as an example, it is equipped with lidar and NVIDIA Orin chip, which means that the company has the hardware conditions to deploy high-level autonomous driving. In this case, with the help of OTA technology, the automatic driving software of the whole vehicle is continuously upgraded in the later period. Consumers can theoretically unlock high-level autonomous driving in a growing number of usage scenarios. It is expected that foreign-funded smart electric vehicles will be fully equipped with lidar in 2-3 years, and this time period is the golden development period of domestic smart electric vehicles.
summary
Under the influence of the epidemic and lack of cores, foreign chip companies are likely to give priority to supporting multinational auto giants. For domestic computing power chips, this is an excellent opportunity to develop large computing power chips.
A good smart car experience must be the reason for consumers to pay. Of course, the experience of large computing power chips is still in the early stage of development. The addition of domestic manufacturers such as Huawei, Horizon, and Black Sesame increases the chance of localization. However, whether the company’s chip computing power can keep up with the update iteration cycle of domestic car companies will be a test for major chip manufacturers. In addition, it is not the various parameter data given by car companies that ultimately judge the quality of its autonomous driving performance, but the real user driving experience. This year, the new car-building forces, Huawei Wenjie, Weilai ET7 and other new models are launched, we will see more real evaluations from users.