Driven by the wave of "new four modernizations", the transformation and upgrading of the automotive industry has been accelerated, and the role of the car itself has gradually developed from a simple mechanical product to a cutting-edge electronic intelligent terminal, and the design architecture of intelligent vehicles is also undergoing major changes, from the traditional distributed ECU architecture to the centralized architecture of domain controllers.
Under this trend, the innovation of smart cars has begun to revolve around chips, AI, electronic architecture and so on. The chip computing power platform has become one of the key factors of change, and the degree of chip integration is getting higher and higher, and the computing power is getting larger and larger, which has become the basis for the development of intelligent vehicles to centralized architecture. The computing power of a single vehicle has also jumped from a few T of computing power a few years ago to the order of thousands of T, and promoting the development of automatic driving through the improvement of computing power has become a "recognized" way in the industry, and automatic driving has entered the era of computing power competition.
At the recent media communication meeting held by Black Sesame Intelligence, Yang Yuxin, CMO of Black Sesame Intelligence, introduced the current situation and importance of the core chip computing power and hardware competition in the current automotive market under the opportunities and development trends of the current automotive market.
Black Sesame Intelligent CMO Yang Yuxin
Yang Yuxin said: "With the evolution of automotive electronic and electrical architecture and the forward development of intelligent vehicles, the support of large computing power core chips is needed. Whether industry manufacturers can grasp the core computing chip platform is a key factor in grabbing the development of the next generation of intelligent vehicles. ”
But this is definitely not a simple chip accumulation and TOPS hash rate competition. Yang Yuxin stressed that from the perspective of chip performance, automatic driving cannot be simply viewed from the value of TOPS, but should be a comprehensive computing power. Autonomous driving is a balanced system supported by comprehensive computing power, which supports different sensors and requires more complete computing power, including CPU, GPU, NPU, DSP and other parts of the comprehensive computing power. Therefore, how to make the computing power of the system more balanced, how to make the chip and software more closely integrated, is the more important focus of the current track.
The strong competitiveness of Black Sesame Intelligence
Founded in 2016, Black Sesame Intelligence, starting from image processing, does artificial intelligence, vision, automatic driving, and large-scale integrated circuits, and has achieved a series of achievements so far. Up to now, the company has created two core IP (vehicle specification image processor NeuralIQ ISP, vehicle specification low-power neural network acceleration engine DynamAI NN), launched the Huashan series of 2 generations of 4 high-performance intelligent driving computing chips (A500/A1000/A1000L/A1000Pro), defined the Shanhai artificial intelligence development platform, and is also the first in China to collect functional safety expert certification, functional safety process certification, AI chip company for product conformity certification.
Products and ecology
Yang Yuxin introduced that the A1000 Pro chip has reached 106TOPS (INT8) in computing power, supports 16 high-definition camera inputs, has high performance, low power consumption, safety and reliability characteristics, can support L3/L4 high-level automatic driving function, and is currently the most powerful domestic automotive standard-level automatic driving chip.
Behind the performance of the chip is so strong, it is the highlight of its advantages in self-developed core IP. Dr. Deng Kun, vice president of black sesame intelligent application engineering, said that one is a self-developed neural network processor NPU, which has 58TOPS (INT8) computing power, and a single chip can process about 10 or more camera data, so that the vehicle can simultaneously process forward-looking, surround view, including side-view multi-camera input information, which can better help cars to monitor and identify objects or lane lines, vehicles, traffic lights and other information, and can achieve a better full-featured automatic driving scene recognition and landing The other is is the ISP, that is, the digital image processor, the image information through the visual sensor to the intelligent driving vehicle, through the digital image processing, can make the noise lower, the dynamic balance is better, so that the vehicle can see more clearly in terms of vision.
Dr. Kun Deng, Vice President of Intelligent Application Engineering of Black Sesame
The two self-developed core IPs add the advantages and competitiveness of Black Sesame Intelligence in chip design.
In addition, in addition to having a complete product system including car end, road end, for different solutions from L2 to L2+ to L3, L4 level, Black Sesame Intelligence has also been committed to creating an open ecosystem. Yang Yuxin said that the autonomous driving chip is relatively speaking, as a large chip, to run the operating system, a variety of applications and algorithms, which determines that it needs to have an open ecological attribute to connect upstream and downstream customers.
Deng Kun said that the cycle and speed of hardware development is slower than that of software, the frequency of algorithm iteration may be measured in weeks or months, the frequency of software iteration may be measured in months or quarters, but the iteration cycle of hardware controllers or hardware devices and chips is measured in quarters or years. Therefore, the current OEMs hope to meet the needs of customers faster, but there will be a hardware development speed that is far from keeping up with the development speed of software or algorithms.
In this regard, it is currently commonly used in the industry to decouple soft and hard. Deng Kun said that Black Sesame Intelligence has truly decoupled software and hardware, proposed a comprehensive open system of soft and hard decoupling, provided different levels of tool chains, and assisted customers to do the transplantation of algorithms and models. This way is also a good compliance with the trend of "software-defined car", the current research and development of hardware is usually the use of hardware burial, computing power embedded in the form of relatively high computing power, because a full-featured product may be a one-year or two-year controller, but its software algorithm can use software and hardware to uncouple independent of hardware development, can continue to use the form of rapid and sustainable iterative development, to continue to adjust the software and algorithms according to customer needs, to achieve more functions.
On the whole, Black Sesame Intelligence not only has a complete full product chain of chips, which are from low-end, medium-matched to high-end chip capabilities, but also has full-stack algorithm capabilities, which is equivalent to providing chips at the same time, but also can provide automatic driving algorithms and intelligent network related algorithms to help customers quickly land and build their data and ecological closed loops.
According to Deng Kun, the company now has a large number of ecological partners in China, including ecological partners in the fields of algorithms, software, maps, positioning, and some hardware partners, and at the same time has carried out mass production cooperation with a number of OEMs in China, and there are also related mass production projects and front-end mass production projects that have been signed, and it is expected that there will be a large number of shipments next year to the next year.
Cross-border integration of teams
Behind the excellent results, it is natural to avoid the advantages of the team.
Yang Yuxin said that we are a team that can completely gather more than 20 years of experience in the automotive industry and more than 20 years of experience in the chip industry in this field. As we all know, many of the current innovation and entrepreneurship are in cross-border integration, but in fact, the integration of cross-border teams is actually very difficult.
Black Sesame Intelligence is a good combination of these two points, the core team established in 2016 includes leading chip companies such as Arm, Qualcomm, Marvell, etc., as well as team backgrounds in the automotive industry such as Ford, Bosch, and General Motors, which also constitutes our unique advantage in this field, understanding both cars and chips. Because whether it is the automobile or the chip industry, it takes a very long time to accumulate, and it is not a short period of time to have the opportunity to do this thing.
In addition to having the advantages of cross-border and professional teams, it is not unrelated to unswervingly develop in the direction of identification from the beginning of entrepreneurship. Yang Yuxin told the media that the company decided to use artificial intelligence perception technology as the basis to empower the future of autonomous driving at the beginning of its entrepreneurship, and since 2016, it has been along this direction to the present, and the general strategic direction has not been adjusted too much, and the time nodes from algorithms, IP, chips, and even including the progress of cooperation with car manufacturers are more clearly grasped.
It can be said that words must be done, and deeds must be fruitful.
The change and change of the automotive supply chain
From the perspective of the BOM cost of smart cars and the entire industry chain, chips account for about 40%-50% of the cost of automotive BOM, and this proportion is still on the rise. Combined with the current trend of global core shortage, many automakers have gradually moved closer to chip companies, and even begun to develop core chips by themselves, hoping to further grasp the right to speak and dominate core technologies and software.
In this regard, Deng Kun believes that under the influence of the development trend of the automotive industry and the general environment, the main engine factory now does have its own ideas and ideas for the integration of software and hardware in the future, and for chip companies, it is hoped to cooperate with the perspective of the development of the main engine factory, and more technology has been previously provided by chip companies to Tier1 manufacturers to support the development of the main engine factory, and gradually become the form of cooperation between chip manufacturers and first-level suppliers, and everyone is more developing towards the form of 1.5-level suppliers.
It can be understood that the first-tier supplier and the second-tier supplier each take out their most advantageous part and cooperate with the R&D process of the main engine factory. This is quite different from the original cooperation model, taking domain controllers as an example, the early OEM development model was to designate a first-tier supplier to develop domain controllers, and the OEMs were only responsible for the functional requirements of the entire car, including project nodes without considering the specific hardware and software implementation. Now we see that most of the domestic OEMs develop the model not only designated a first-tier supplier, but also designated software suppliers, hardware suppliers, algorithm suppliers, chip suppliers including system integration suppliers, by four or five suppliers to jointly complete a project research and development. The second-tier supplier has gradually faced the main engine factory from the role of not facing the main engine factory in the past, and has now become a chip supplier to gradually face the main engine factory and join in to discuss the functional requirements.
This trend may be a better opportunity for chip manufacturers, so that they can have more expressiveness, as well as more opportunities to understand the needs and trends of OEMs.
On the other hand, for the phenomenon of self-developed chips of the main engine factory, Deng Kun believes that the main engine factory has a lot of self-developed chip ideas, including planning, but through communication with the main engine factory, it has also been found that the main engine factory's own talent reserves, technical reserves, including industry reserves, may not be able to completely support the landing of a chip from beginning to end, and more importantly, the main engine factory will put this demand more clearly, and more will choose and chip companies to customize chips for now mature technology and mature IP.
It is not difficult to see that the above industry dynamics and trends are promoting the reconstruction of the automotive supply chain ecology. In this process, chip manufacturers seem to have ushered in more development opportunities and opportunities for role selection. "Our core positioning is still Tier2, and the current restructuring of the automotive industry chain comes from the boundaries of the division of labor, but its essence will still follow the logic of business, in the short term, car manufacturers will not directly purchase chips from chip companies on a large scale, but the redefinition of this boundary and role has just begun."
Black Sesame Intelligence hopes to become the leader of automatic driving computing chips, share development opportunities with car manufacturers, and empower car manufacturers to create differentiated products through its own technical advantages, product system, open ecology and flexible business model to help the development of automatic driving.
Write at the end
The past year has been a year of "lack of cores" in the global automotive industry, and the organic crisis has provided opportunities for depots to rebuild their supply chains. In the local supply chain system of domestic car manufacturers, domestic suppliers are playing an increasingly important role.
Yang Yuxin believes that under this opportunity, China's automotive industry chain has opened an opening for domestic car chip manufacturers, but this window cycle will not be too long, about 3-5 years, we must seize the three or five years of time cycle to do our products, and strive to form a domestic supply chain system, once the new supply chain system matures, this alternative window will be closed.
Speaking of the strategic goals and market opportunities of Black Sesame Intelligence in the next five years, Yang Yuxin believes that the opportunities for automotive chip manufacturers in the future come from the demand for a large number of new chips brought about by the changes in the electronic and electrical architecture of the car itself. For black sesame, a very important strategy for the development of the next five years is to quickly expand the product line according to the needs of the intelligent functional improvement of the car, and continuously improve the value of the bicycle. The goal of Black Sesame Intelligence is to create core technologies and products under this wave of automotive intelligence, truly promote the rapid commercialization of autonomous driving, and lead the development of the industry.
Source: Semiconductor Industry Watch