Completed all mass production certification in two years, Black Sesame Intelligent high-computing power autonomous driving chip will be put into vehicles within this year

The countdown to the introduction of domestic high-computing power chips into automobiles has begun!

Just this week, domestic autonomous driving chip company Black Sesame Intelligent announced that its Huashan II A1000 chip has entered large-scale production and has begun to be shipped continuously to industry customers, and will be mass-produced and installed in vehicles within this year.

Looking at the entire mass-produced autonomous driving market, L2 autonomous driving has long become the standard for new cars priced at RMB 100,000. By 2022, people have higher expectations for autonomous driving functions, such as navigation-assisted driving, urban autonomous driving, and autonomous valet parking. High-computing autonomous driving chips have naturally become an indispensable condition for realizing these functions.

However, in the global autonomous driving chip market, although there are mass-produced high-computing power autonomous driving chip solutions, the cost is relatively high, and it is difficult to make them standard equipment in models priced above 300,000 yuan.

Obviously, the industry lacks a highly cost-effective and high-computing-power chip solution.

In response to market demand, Black Sesame Intelligence, a domestic autonomous driving chip company, released the Huashan-2 A1000 autonomous driving chip in 2020. After two years of software and hardware polishing and verification, A1000 will become the first high-computing autonomous driving chip in China to complete all mass production certifications.

So, what is the verification status of the A1000 chip before mass production? Recently, Chedongxi had an in-depth exchange with Yang Yuxin, CMO of Black Sesame Intelligence, and found the answer to the question.

1. The demand for L2 autonomous driving is high, and even a 100,000 yuan car can be equipped with a high-computing chip

Nowadays, mass production of autonomous driving has entered a period of almost daily iteration. After various advanced L2 functions such as navigation-assisted driving, memory parking, and urban L2 have been mass-produced, the mass production of autonomous driving industry is facing a new problem - insufficient computing power.

To achieve L2 autonomous driving on complex urban roads, it is necessary to record parking routes of several floors and hundreds of meters in the parking lot, and data from multiple cameras, multiple millimeter-wave radars, and even lidars are essential. At this time, the amount of data generated by sensors has increased exponentially, and the performance of traditional low-computing autonomous driving chips is no longer sufficient.

As the demand for L2 autonomous driving and computing power is growing explosively, autonomous driving chip manufacturers around the world have seen the future of high-computing-power chips.

Nvidia, which had an early entry into the market, has achieved mass production and installation of the Orin autonomous driving chip this year after releasing it in 2019. The first batch of models equipped with it are NIO ET7 and Zhiji L7.

However, looking through the entire market, there is no other mass-produced high-computing autonomous driving chip solution except NVIDIA. Moreover, NVIDIA's high-computing solution is only installed in models above 400,000 yuan, and the important reason behind this is that the cost is too high.

Such market situation directly leads to the difficulty of popularizing advanced L2 autonomous driving. How to make mainstream models with a price range of 100,000 to 300,000 yuan also equipped with high-computing chips is the future development direction of the mass-produced autonomous driving industry.

Faced with this market gap, the opportunity to mass-produce high-computing-power autonomous driving chips naturally falls to domestic manufacturers.

The good news is that 2022 is the first year that domestically produced high-computing-power autonomous driving chips will be installed in vehicles, and we have now entered the countdown stage.

2. All mass production certifications completed and delivery in June

Not long ago, Shan Jizhang, founder and CEO of Black Sesame Intelligence, said at the China Electric Vehicle 100 Forum: "The Huashan II A1000 autonomous driving computing chip developed by Black Sesame Intelligence is fully mature in terms of functional safety, information security, and reliability. It will be mass-produced and installed in vehicles this year to achieve L2-L3 autonomous driving functions."

The latest information shows that the Huashan-2 A1000 chip has begun mass production and is scheduled to be delivered from June this year.

▲The left and right are the Black Sesame Huashan No. 2 A1000L and A1000 chip

This means that Heizhima Intelligence has seized the opportunity to install high-computing-power autonomous driving chips in vehicles, and A1000 will become the first high-computing-power autonomous driving chip in China to complete mass production certification.

In fact, as early as June 2020, Heizhima Intelligence had successfully taped out and released the A1000 chip. This chip is built based on the 16nm process and its AI computing power can reach up to 58TOPS (INT8).

Since then, the A1000 chip has entered the intensive mass production verification stage.

Yang Yuxin, CMO of Heizhima Intelligence, introduced that during the two-year mass production verification phase, the most challenging part is the automotive-grade certification of the chip.

Within Heizhima Intelligence, there is also a team responsible for vehicle regulation certification. This team plays the role of project management and coordinates the cooperation of various departments such as chip design, underlying software, middleware, and upper-level software.

In June 2019, before the release of the A1000 chip, Heizhima Intelligence had obtained security team certification; in July 2020, the A1000 chip obtained ASIL-B ready functional safety product certification; and in June last year, it obtained ASIL-D functional safety process certification again. At the same time, Heizhima Intelligence has also obtained ISO9001:2015 and ASPICE CL2 certifications. In early 2022, the A1000 ASIL-B functional safety product certification was completed, making it the first domestically produced high-computing power chip in the industry to complete this certification. In addition, with the AEC-Q100 Grade 2 certification, the A1000 chip is qualified to "work with a certificate."

▲Heizhima Smart obtains automotive-grade functional safety certification

In addition to passing multiple automotive certifications, Heizhima Intelligence has also completed the transplantation and verification of mainstream industry operating systems, deep learning development tool chains, professional middleware software for autonomous driving, and verification support for multiple AUTOSAR AP CPs.

In addition, Heizhima Intelligence also provides a large number of sample codes suitable for different autonomous driving functions and scenarios, such as GPU applications for 3D surround rendering, algorithm applications for forward and lateral fusion perception, H.264/H.265 encoding and decoding programs for driving recording, and fisheye perception for automatic parking.

At the system level, Heizhima Intelligence provides diagnostic and safety programs, supplemented by various production line tools, to provide maximum convenience for rapid mass production for industry customers with different needs.

It takes two years to complete all the work required for mass production of chips. This time span is the most basic operation for mass production of autonomous driving chips. The A1000 chip, which is qualified to "hold a certificate to work", is ready for mass production.

3. New car will be launched this year. A team of dozens of people will be stationed in the factory to ensure the mass production of the first project

When talking about the launch time of mass-produced cars, Yang Yuxin said that the new cars equipped with Black Sesame Intelligent A1000 series chips are planned to be mass-produced and launched this year.

In addition, Heizhima Intelligence has obtained designated projects from many automobile companies, covering multiple models such as passenger cars and commercial vehicles. Other models equipped with A1000 chips will also be released this year.

Chedongxi learned that from now until the new car is launched, Heizhima Intelligent's engineers will continue to work with vehicle manufacturers.

Yang Yuxin said that for every vehicle mass production project, Heizhima Intelligence will set up a dedicated service team.

Regarding the cooperation model with vehicle manufacturers, Yang Yuxin said that Heizhima Intelligence provides customers with "nanny-style technical support". Engineers will be stationed in the factory for a long time to work together with the engineers of the vehicle manufacturers, and continuously improve products according to the requirements of the vehicle manufacturers' customers, thereby accelerating the progress of mass production.

▲Black Sesame Smart Headquarters

Looking at the overall R&D strength of Heizhima Intelligence, it now has nearly 700 employees, more than double the number a year ago. More than 90% of them are engineers and technicians, focusing on technical support to fully cooperate with the commercialization of A1000.

Yang Yuxin introduced that this has a lot to do with the good business growth achieved by Heizhima Intelligence in the past year and the increasing demand for software research and development.

Among many domestic self-driving chip startups, Heizhima Intelligence has become the first company that can mass-produce high-computing self-driving chips, thanks to its precise product positioning and strong technical strength. This has also become the only way for domestic self-driving chip companies to win more orders and win more car companies.

4. Two core IPs developed independently, the next generation of high-computing chips will be released this year

From a technical perspective, Heizhima Intelligent's self-developed DynamAI NN neural network processor and NeuralIQ ISP image signal processor are the core of autonomous driving chips.