The "volume" of the automotive chip track may have exceeded everyone's expectations. The competition of pure TOPS computing power has been overturned. How to make car companies use it, use it well, and use it well is already a new trend.
In fact, in 2018, when car intelligence just started, car-machine systems like Banma Zhixing only changed the human-computer interaction experience of the traditional cockpit from the software level (from functional phones to smart phones). ADAS visual perception system solutions like Mobileye (before EyeQ5) are only entry-level for assisted driving.
In the view of Gaogong Intelligent Automobile Research Institute, the automotive chip track has gone through several development cycles. In the 1.0 era (represented by Qualcomm 8155 that will be launched in 2020), the smart cockpit has entered a node of hardware change; Represented by the Nvidia Orin of the car), intelligent driving has entered the node of hardware transformation.
In the 3.0 era, there are two paths. First, when the overall cost reduction pressure in the automotive industry is increasing sharply, high-performance + cost-effective solutions have become the main force in the market; second, in line with the upgrade of the vehicle's electronic architecture (centralization, cross-domain integration), the solution architecture of the computing platform has been proposed new requirements.
For example, last year, Nvidia and Qualcomm launched their first central computing platform solutions (Thor, Ride Flex), which focused on super-large computing power + cross-domain integration. On the one hand, they provided car companies with options for technical planning in the next few years; On the one hand, it is chip manufacturers competing for cross-domain opportunities. For example, Qualcomm is better than the cockpit, and Nvidia is better than smart driving, but any segmented track is not enough to support scale.
Whether it is smart driving, smart cockpit, central computing control and other fields, chips, as the basic computing platform, have not only been the focus of public opinion in recent years, but have truly become the "entry point" for automakers and parts giants to enter the next competitive cycle. ticket".
For other automotive chip manufacturers, there are many tests on how to form a differentiated competitive advantage with Nvidia and Qualcomm.
This week, as an industry-leading research and development enterprise of automotive-grade autonomous driving computing chips and platforms, Black Sesame Smart officially launched the company’s positioning upgrade (upgraded from "leader of autonomous driving computing chips" to "leader of smart car computing chips") At the same time , it released a new product line - Wudang series smart car cross-domain computing platform and its first chip C1200.
"When the intelligent driving technology begins to mature, the Chinese market is actively exploring its own technical route. We hope to bring disruptive and innovative products and technologies to the industry to meet the needs of the local market." Shan Ji, founder and CEO of Black Sesame Smart According to Zhang, "Keep the right path and make surprise soldiers" is the answer given by the company.
For a company focused on automotive-grade chips, scale is more critical than in other industries. Whether it is 70 to 80 million new cars delivered each year in the world or more than 20 million new cars in the Chinese market, the ceiling of a single application is obvious.
In this regard, Black Sesame Intelligence has given a "three-step strategic positioning" plan. Among them, continue to focus on autonomous driving computing chips and solutions, realize the commercialization of products, and form a complete technical closed loop, which is regarded as "keeping the right path" .
For example, the Huashan series A1000 chip can already well support the latest Transformer-based BEV algorithm, which can fully meet the needs of the next L3 and below autonomous driving. At present, this series of chips is in mass production (all vehicle-level certifications have been completed), and mass production will be launched this year.
In addition, Black Sesame Intelligence has the ability to mass-produce full-stack perception algorithms, can provide customer algorithm customization services, support third-party algorithm transplantation, and multiple algorithm delivery methods and other business models. This is also in line with the multi-level needs of the current automotive industry. At the same time, Black Sesame Intelligence has successively released the Shanhai artificial intelligence development platform and the Hanhai autonomous driving middleware platform. The mature tool chain and middleware system support rapid mass production.
But this is not enough.
Because the automotive industry is entering a critical cycle of electronic architecture upgrades, with diversified demands, centralized architecture, and the pressure of cost games, for chip manufacturers, "surprising soldiers" are needed.
In the view of Gaogong Intelligent Automobile Research Institute, the first surprise soldier dispatched by Black Sesame Intelligence is "a cost-effective driving and parking integrated intelligent driving domain control solution within 3,000 yuan" .
From the perspective of Black Sesame Intelligence, the automotive industry, especially the smart driving track, has shifted from blindly pursuing high-index or high-performance functional modules to taking into account performance and cost performance. Cost-effectiveness brings the confidence to develop new technologies, and is an important link to promote the large-scale and sustainable development of smart cars.
For example, as an upgraded version of the entry-level driving L2, NOA (automatically assisted navigation driving, from point A to point B) and integrated driving and parking solutions are the highlight combination functions of many car companies in recent years. At the same time, by increasing the perception redundancy of lidar, some car companies are also promoting the landing from high-speed scenes to urban scenes.
However, according to the monitoring data of Gaogong Intelligent Automobile Research Institute, based on the data of the Chinese market in 2022, the average price of NOA's pre-installed standard models is 397,400 yuan; in comparison, L2 (excluding L2+) pre-installed standard The average price of the model is 246,800 yuan. This means that the popularization of high-end and high-low speed combined intelligent driving functions requires rapid cost reduction.
"Providing cost-effective products under the premise of ensuring performance" is also regarded as the key to the scale-up of high-end assisted driving pre-installation standard equipment in the next 3-5 years.
The solution given by Black Sesame Intelligence is to control the BOM cost of the integrated domain controller supporting 10V (camera) NOA function within 3,000 yuan, support physical computing power of more than 100T, and help car companies solve cost pressures.
As the first domestic chip platform in China with a single chip that supports the domain controller integrated with parking and parking, the Huashan No. A certain degree of improvement is considered "system-level cost reduction."
So far, Black Sesame Smart has cooperated with Dongfeng, Jiangqi Group, Sany Group, China FAW, Bosch, SAIC, SAIC-GM-Wuling, Dongfeng Yuexiang, Zhongke Chuangda, Asia Pacific, Baolong Group, Jingwei Hengrun, Junlian Zhixing , Suo Tuo Ruian, Lianyou Technology, etc. have carried out a series of commercial cooperation on L2, L3 ADAS and autonomous driving perception system solutions.
At the same time, Black Sesame Smart also dispatched the second "Advanced Soldiers", the Wudang series smart car cross-domain computing platform and its first chip C1200. This is considered to be based on the development trend of automotive electronic and electrical architecture. The company expanded its product line to cover more computing nodes in the car, forming a combination of multiple product lines, behind which is the increase in the value of single-car products.
"Whether the value of a single car (multiple chips that don't need to be used in the car) can reach 1,000 US dollars," this is the goal set by Shan Jizhang for Black Sesame Smart, because for the semiconductor industry, "shipment" is the basic indicator , otherwise it will not be profitable.
However, how to build a cross-domain computing platform that can truly meet the large-scale landing of the Chinese market, simply "copying" the NVIDIA or Qualcomm model is obviously not the best policy. The answer given by Black Sesame Intelligence is: start from the actual needs of users.
For car companies, being able to support the large-scale implementation of smart electric vehicles is an urgent need for the upstream of the supply chain.
Facing cross-domain computing scenarios, the Wudang series, as the industry's first smart car cross-domain computing chip platform, can perfectly cover the needs of multiple different domains inside smart cars such as cockpit and smart driving, and has the ability to integrate multiple domains. Behind this is the family platform of Black Sesame Smart, which ensures the best balance of computing power, power consumption and cost of the chip.
"In terms of chip design, an innovative architecture is needed to better integrate multiple functions from different domains in a practical, easy-to-use, and easy-to-use way." According to Ding Ding, vice president of Black Sesame Smart Products, This key architectural innovation carries 7 types of computing power upwards, and manages the corresponding data efficiently and flexibly through this infrastructure.
For example, the Wudang series product line based on ESDE, a self-developed low-latency, high-throughput data exchange architecture, clearly aims at the massive L2+ level intelligent driving and converged computing application market. With innovative fusion architecture and heterogeneous isolation technology, the Different computing power can be matched and combined according to different scenarios, as well as different specifications and safety requirements, to support the flexible development of automotive electronic and electrical architecture.
"Extreme cost performance" is the main label that Black Sesame Smart put on Wudang C1200. "Many companies in the market can only do a good job in the ability of a single-domain computing platform. Being able to span multiple domains has a high threshold. On this basis, it is even more difficult to emphasize cost performance." Shan Jizhang said frankly.
In terms of performance parameters, C1200 is based on the 7nm advanced process, using the car-grade high-performance CPU core A78AE (performance up to 150KDMIPS) that supports lockstep, and the car-grade high-performance GPU core G78AE, providing powerful general computing and general rendering computing power ; Self-developed DynamAI NN car-grade low-power neural network acceleration engine, supporting NOA scenarios;
Built-in mature high-performance Audio DSP module and a new generation of self-developed NeuralIQ ISP module with online processing of 1.5G pixels per second; provides the industry's highest MCU computing power of 32KDIPMS; can simultaneously process the input of more than 12 high-definition cameras, and supports high-speed MIPI .
In terms of peripherals, C1200 supports access and forwarding of multi-channel CAN data, supports Ethernet interface and supports all commonly used display interface formats, supports multi-screen output, and multi-channel 4K capability; supports dual-channel LPDDR5 memory particles, which can meet Bandwidth requirements after cross-domain integration.
At the same time, the C1200 pioneered the hard-isolated independent computing subsystem, independent rendering, and independent display, which meets the requirements of high security and fast startup of the instrument control screen. At the same time, this subsystem can also be flexibly applied to computing scenarios that require independent systems, such as autonomous driving and HUD head-up display.
The built-in security module that supports ASIL-D level Safety Island, National Secret Level 2 and EVITA full can meet the reliability requirements of the highest safety level of the vehicle regulations.
More importantly, C1200 can flexibly support various current and future architecture combinations in the industry. A single chip can meet requirements including CMS (electronic rearview mirror) system, driving and parking integration, vehicle computing, infotainment system, smart headlights, cabin, etc. Cross-domain computing scenarios such as internal perception systems have great value and cost advantages.
"Black Sesame Smart is committed to providing customers with the best products through subversive technological innovation." Shan Jizhang said that the C1200 is expected to provide samples in 2023, and Black Sesame Smart will cooperate with car manufacturers, Tier1, software and hardware strategic partners, Create innovative benchmark products based on C1200.
As for the future trend of market competition, Shan Jizhang judged, "Whether it is for the semiconductor or automobile industry, it is not realistic for one company to dominate, but in the end there will not be too many leftovers. At the same time, considering the inherent nature of the chip industry In the time period from R&D to mass production, the chances of new entrants will only get smaller and smaller.”
At the same time, policy support for China's domestic market is also in place one after another. In the view of Gaogong Intelligent Vehicle Research Institute, this is a huge benefit to China's smart car chip market.
This week, the Science and Technology Department of the Ministry of Industry and Information Technology publicly solicited opinions on the "Guidelines for the Construction of the National Automotive Chip Standard System (2023 Edition)" (Draft for Comment). The "Construction Guide" emphasizes the increasingly prominent role of automotive chips in the development of the automotive industry. This is another key top-level design after China's leadership in electrification.
The "Construction Guide" puts forward the goal that by 2025, more than 30 key standards for automotive chips will be formulated; There are standard supports.
At the same time, these standards will cover general requirements such as environment and reliability, electromagnetic compatibility, functional safety and information security, as well as key product and application technical requirements such as control chips, computing chips, memory chips, power chips and communication chips.