China and Africa News, April 26, in the era of intelligence, the automotive industry is undergoing unprecedented changes. With the rapid development of intelligent driving technology, the automotive industry has turned a new round of competition to autonomous driving. In the new models released by car companies recently, assisted driving has become a standard feature, and the ability to achieve higher levels of autonomous driving has become a new selling point.
It is understood that from January to March this year, among the models released by many car companies, the demand for the autonomous driving computing power platform has reached the level of 1000 TOPS. However, there are only a few foreign chip manufacturers that can achieve mass production of computing platforms with large computing power. The domestically-produced car-level large computing power platform is also in urgent need of breakthrough.
On April 19, Black Sesame Intelligence released its new generation of high-performance automotive-grade autonomous driving computing chip Huashan No. 2 A1000 Pro on the first day of the 2021 Auto Show. The chip can reach 106 (INT8) -196 (INT4) TOPS computing power, adopts heterogeneous multi-core architecture, 16-core Arm v8 CPU, 16nm process, and the typical power consumption is only 25w. This is currently the domestic autonomous driving computing chip with the highest computing power. It can support high-level autonomous driving functions, seamlessly connecting from parking, urban interiors to high-speed scenes.
But computing power anxiety is obviously not the only competition point in this global automotive revolution. Entering 2021, more and more technology giants such as Baidu, Xiaomi, Didi, and Huawei are pouring into the track of the automotive revolution. New cars have also entered a new stage. The original vehicle technology is still the key, but more focused on electronics. The new propositions of electrical architecture, intelligent driving, and intelligent cockpits urgently need to be supported by a new industrial ecology.
“Our in-depth cooperation with car manufacturers on R&D is almost like a company,” said Shan Jizhang, founder and CEO of Black Sesame Intelligence Technology (hereinafter referred to as “Black Sesame Intelligence”).
How to treat computing power anxiety in a correct way?
With the global data explosion and the unfolding of the artificial intelligence revolution, the growth of computing power can no longer meet the demand for data growth. According to IDC statistics, the global demand for computing power doubles every 3.5 months, far exceeding the current growth rate of computing power.
In the global competition for autonomous driving, the competition around computing power is the focus of almost everyone, and the pursuit of large computing power in chips has almost become an important selling point for various car companies to launch smart new cars.
Just a week before the auto show, Nvidia just released its first data center CPU, and thus gathered the “three-core” combination of CPU, GPU, and DPU. In addition to facing the high-performance computing market, it also points directly to the field of autonomous driving. Introduced a new generation of NVIDIA DRIVE Atlan for self-driving cars, the processor performance is up to more than 1,000 trillion operations per second (TOPS).
In this regard, Shan Jizhang believes that this requires an overall look at the car manufacturers’ requirements for chips. “In fact, computing power is only a part of the middle. It is a heterogeneous SOC and a system. It needs to look at power consumption and performance. , Comprehensive consideration of three aspects of cost.”
How do you understand it in detail? A single chapter explains, for example, the main calculations in the car may be through neural networks, and there are various sensing systems, including various calculations such as lidar, ultrasonic radar, millimeter wave radar, and cameras. It does require a lot of computing power, which is why everyone now regards chip computing power as a major parameter. But in fact, the parameters of the chip are much more complicated. In addition to the functional safety of the car itself, it also puts the path planning, decision-making, and control algorithms in the entire system on the chip, so the CPU , Sensor fusion and path planning algorithms have high requirements, but also must consider reducing power consumption, which is a very complex issue.
There is also a requirement for the chip to be easy to use. Shan Ji Zhang pointed out that this is also the reason why Black Sesame Intelligence cooperated with the Huashan series of autonomous driving chips this time and also released the Shanhai artificial intelligence development platform. According to reports, the platform has more than 50 AI reference model library conversion use cases to help reduce the threshold for customers’ algorithm development; it can achieve comprehensive optimization of QAT and post-training quantification to ensure the accuracy of algorithm models; it supports dynamic heterogeneous multi-core task allocation, and also Support customer custom operator development, complete tool chain development kit and application support, which can help customers quickly transplant models and deploy the integrated process of landing.
Shan Jizhang said that it is becoming more and more critical that the automotive smart application experience is made more detailed to meet the market demand, which requires chip companies to provide customized support for downstream application needs, which is a particularly important point. “The chip company must understand the application scenarios, look at various applications from a system perspective, and not only implement these applications, but also abstract the commonality from it.”
Getting together to “build a car” is not a bad thing, the auto industry is changing the rules of the game
Regarding the past six months, major technology companies have officially announced that “car making” has become crowded into the tide of automobile intelligence. Shan Jizhang believes that “this is a good thing”, and it will be possible for industry chain companies including chip companies. Benefit from the resulting series of industrial ecological innovations.
“Now the transformation of the entire automotive industry has just begun. With the influx of major technology companies, everyone starts with their own strengths and promotes changes in the entire industry from different directions. This is very similar to the situation when the smartphone was just emerging around 2010. .” Shan Jizhang said, and in his opinion, the automotive market will undoubtedly be much larger than smart phones.
For the huge market of intelligent driving, it is a very good thing that many players and manufacturers have entered in the early stage. “First show that this direction is right. At the same time, each company has its own accumulation and strengths. Everyone uses different methods and different angles to promote the market, and eventually find the optimal model. At the same time, the market is large enough, Not one or two companies can take it all.”
According to McKinsey’s forecast, by 2030, the order value of travel services based on autonomous driving will reach approximately US$260 billion, and by 2040 it will reach approximately US$940 billion. Such a huge market can accommodate more players. With the entry of more and more technology giants, a fully competitive market structure is taking shape.
In addition, Shan Ji Zhang also specifically mentioned that the advancement of technology, especially the development of artificial intelligence, has promoted the rapid application of the latest process technology to automobiles, and at the same time has produced a new business model, which is “software-defined cars.” For example, the subscription model launched by Tesla, which embeds high-performance hardware in the early stage, allows the software to be upgraded online to achieve the increase of car functions and performance. Of course, these require the support of the underlying hardware to provide sufficient computing power and redundancy. For chip companies, it will bring more new opportunities.
Shan Ji Zhang said that in the face of such a huge market trend, as a chip manufacturer, what you need to focus on is what kind of technical preparations to make and how to continuously adapt to the fast-changing upstream and downstream industries and business models.
In addition, Black Sesame Intelligence also announced a strategic cooperation with Dongfeng Design and Research Institute and Dongfeng Yuexiang during the auto show, forming a close partnership from joint development of cutting-edge technologies to mass production applications. Black Sesame Intelligence will carry out in-depth cooperation with Dongfeng Design and Research Institute and Dongfeng Yuexiang in areas such as autonomous driving, vehicle-road collaboration, and smart logistics. While carrying out joint technological development, it will also build on Black Sesame intelligent autonomous driving computing chips. The vehicle-level intelligent driving platform further accelerates the commercialization of autonomous driving.
According to reports, Black Sesame Intelligent Technology Co., Ltd. is an industry-leading research and development company for automotive-grade AI computing chips and platforms for autonomous driving. It focuses on high-tech research and development in technical fields such as large computing chips and platforms. Vehicle-road collaboration solutions, including autonomous driving perception computing chips and autonomous driving computing platforms based on vehicle-level design, learning image processing, and low-power precision perception, support the rapid industrialization of related product solutions in the autonomous driving industry chain. The company established R&D centers in Silicon Valley and Shanghai in 2016, with offices in Chengdu, Shenzhen, Wuhan, Chongqing, and Singapore. It currently has nearly 400 employees. The core teams are all from Bosch, OV, Nvidia, Ambarella, Microsoft, Top companies in the industry, such as Qualcomm, Huawei, and ZTE, have an average of 15+ years of industry experience.