How is Heizhima Intelligence developing in the field of intelligent driving chips?

With the rapid transformation of the automotive industry towards intelligence and electrification, autonomous driving technology has become an important direction for future development, and the in-vehicle SoC (system-level chip ), which is the core technology of autonomous driving, has rapidly emerged. As a leading domestic Tier 2 chip supplier, Heizhima Intelligence has gradually stood out in the intelligent driving chip market with its strong technological innovation, rich ecological empowerment and mass production advantages .

Black Sesame Intelligence's core products mainly include two series of automotive-grade SoCs: the Huashan series and the Wudang series. These two series cover the needs of autonomous driving at multiple levels. At the same time, Black Sesame Intelligence actively deploys its ecosystem, and through its self-developed development tool chain and middleware , it has formed a full-stack capability from chips to solutions, serving automotive OEMs and Tier 1 suppliers. With its self-developed IP and complete SoC design capabilities, Black Sesame Intelligence has taken the first-mover advantage in technology, ecology and market, and has been widely recognized in the market competition for smart driving chips.

Technical advantages: strong support from self-developed IP and full-stack design capabilities

1. Advantages of self-developed core technologies

In the smart driving SoC market, Black Sesame Intelligence stands out with its unique self-developed capabilities. Self-developed IP cores give the company the technical initiative in the field of automotive-grade SoC. Black Sesame Intelligence not only has rich independent intellectual property rights in SoC chip design , but also has achieved technological breakthroughs in core IP cores. For example, the company's A1000Pro series of high- computing SoC chips have performed well in L2+ autonomous driving scenarios, with a computing power of 106TOPS at INT8 precision, making it one of the few companies in China with this level of technology.

The independent research and development of core technologies not only improves the technical flexibility of Black Sesame Intelligence, but also significantly reduces the cost and iteration cycle problems caused by outsourcing IP. Compared with the products of foreign manufacturers (such as Mobileye , NVIDIA , etc.), Black Sesame Intelligence's self-developed IP core is more market-adaptable, especially in the complex Chinese autonomous driving market environment, and can better meet the needs of local customers. At the same time, Black Sesame Intelligence's independent IP enables its products to maintain relatively independent and autonomous technical iteration capabilities in the face of instability in the global chip supply chain, reducing dependence on external IP suppliers.

2. Product iteration and technology roadmap

Black Sesame Intelligent's technology research and development roadmap is highly continuous and forward-looking. After the successful mass production of the A1000 series, the company continued to layout the high-computing power and cross-domain computing SoC market, and plans to mass-produce the C1200 cross-domain SoC chip in 2025. This chip can not only play a role in autonomous driving scenarios, but also integrates the functions of smart cockpit and in-vehicle and out-of-car communication, becoming a core component of the smart car multi-domain fusion solution.

Another highlight of the technology roadmap is that Black Sesame plans to mass-produce the A2000 SoC chip in 2026. The goal of this chip is to support L4 and above autonomous driving. It uses advanced 7nm process technology and its computing power will far exceed the mainstream products on the current market. Compared with the current competition of most manufacturers at the L2+ and L3 levels, Black Sesame tries to take the initiative in the future L4 market by deploying high-end autonomous driving chips in advance . This forward-looking product iteration planning enables Black Sesame to not only meet current market demand, but also cope with the challenges of future technological development.

3. Technological innovation of ISP and NPU

In the architecture of SoC chips, the image signal processor (ISP) and the neural network processor (NPU) are two core modules. Black Sesame Intelligent's NeuralIQ ISP technology provides strong support for image processing in multi- camera environments , especially in complex lighting conditions (such as night, rainy and foggy weather, etc.), and can still ensure high-quality image output. The key role of this ISP architecture in intelligent driving is to ensure that the vehicle can accurately perceive the surrounding environment in extreme environments and avoid decision-making errors caused by blurred vision.

At the same time, Black Sesame Intelligence's Dynam AI NPU architecture also demonstrates extremely strong computing power. Through its low-power proprietary ASIC structure, DynamAI can efficiently process data from a variety of sensors . Its third-generation DynamAI NN even natively supports the Transformer architecture, which can achieve end-to-end deep learning reasoning in high-level algorithms for autonomous driving . Compared with the NPU architecture of international giants such as NVIDIA, Black Sesame Intelligence's DynamAI has outstanding performance in energy efficiency and adaptability, and is particularly suitable for China's local autonomous driving needs.

Ecosystem empowerment: dual support of tool chain and middleware

1. Shanhai development tool chain: flexible and scalable algorithm support

The development of autonomous driving technology relies on a lot of algorithm development, debugging and deployment. To this end, Heizhima Intelligent launched the Shanhai Development Toolchain to provide developers with all-round support from model training to actual deployment. The toolchain is highly open, supports mainstream deep learning frameworks such as Tensorflow, Pytorch, ONNX, and can automatically optimize the operating efficiency of different algorithms on SoC chips.

In actual applications, the Shanhai development tool chain helps customers accelerate the entire process from algorithm development to product implementation. It provides a wealth of algorithm reference models and conversion use cases, allowing customers to make adjustments based on existing models, reducing the difficulty of algorithm development. Especially for autonomous driving scenarios above L3 level, the Shanhai development tool chain can migrate, quantify and deploy models according to actual scenario requirements, ensuring that the algorithm can run efficiently on high-computing power SoC chips.

Compared with other SoC suppliers, Heizhima Intelligent's Shanhai toolchain not only supports a wider range of algorithm libraries, but also has high flexibility and scalability. For example, when dealing with dynamic heterogeneous tasks in different autonomous driving systems , the Shanhai toolchain can allocate computing resources in real time to ensure the stable operation of the system. This toolchain design concept not only solves technical problems in the development process, but also provides customers with more options when upgrading the system.

2. Hanhai Middleware: Providing flexible support for multi-domain applications

Middleware is the key layer to achieve the collaboration between hardware and software in intelligent driving systems . The Hanhai- ADS P middleware developed by Heizhima Intelligent provides customers with powerful hardware and software interface support. Hanhai middleware has extremely high openness and compatibility, can support the needs of various autonomous driving systems, and help customers reuse the same software applications in different models and hardware configurations, thereby reducing development costs.

Hanhai middleware is not only suitable for L2+ and L3 autonomous driving, but also can play a role in more complex multi-domain applications such as smart cockpits and vehicle-road collaboration. Its open API interface allows developers to easily access the core processing functions of SoC chips and flexibly adjust them according to needs. At the same time, Hanhai middleware also supports multi-sensor fusion processing, which provides important support for the increasingly complex perception requirements in current autonomous driving systems.

Through platform design, Hanhai middleware not only helps customers reduce development time, but also effectively responds to future system upgrade needs. Its high compatibility means that customers can achieve seamless migration on hardware platforms of different suppliers, which is particularly important in the current rapid evolution of the autonomous driving industry. Compared with its peers, Heizhima Intelligent's Hanhai middleware provides higher flexibility and scalability, allowing its SoC products to better meet complex market needs.

Mass production advantage: comprehensive layout from vehicle end to road end

1. Rapid expansion of the vehicle-side market

Black Sesame's A1000 series SoC chips have been used in multiple mass-produced models, including Lynk & Co 08 and Hycan V09, which cover a wide range of market demands from mid-range to high-end. With its powerful computing power and independent intellectual property rights, Black Sesame's SoC products have been widely recognized in the vehicle market. In particular, its A1000L chip, with its 16TOPS computing power, has become a highlight in the L2+ autonomous driving market.