The intelligent upgrade of automobiles has brought strong growth momentum to the automotive storage market, which has become the third largest core sector in automotive semiconductors, second only to computing SoCs and high-performance MCUs. Currently, domestic storage manufacturers are gradually emerging in the automotive storage market, becoming important participants in China's automotive-grade storage market.
However, the trend of integration and convergence of the vehicle's electronic and electrical architecture, the diversified iteration of various intelligent connected functions, and the empowerment of new technologies such as AI big data models also bring new demands and challenges to in-vehicle storage products.
Recently, YEESTOR, a subsidiary of YEESTOR Microelectronics Co., Ltd., attended the China IC Unicorn - Automotive-Grade Chip Summit. Mr. Yuan Ye, head of the automotive electronics market, delivered a keynote speech entitled "The Way of Intelligent Vehicle Data Storage", sharing YEESTOR's profound understanding of automotive storage and cutting-edge application solutions, and explaining in detail how YEESTOR will empower the continuous development of intelligent vehicles.
Automotive SoC empowers the development of intelligent vehicles
The future of automobiles will continue to develop towards intelligence, connectivity, safety, and energy conservation. New applications and functions such as advanced driver assistance systems (ADAS), autonomous driving, and vehicle-to-everything (V2X) are emerging one after another. Technologies such as algorithm chips, millimeter-wave radar, lidar, and new MEMS sensors are developing rapidly, and traditional MCUs are finding it difficult to meet the development needs of ADAS systems.
So, how can chip manufacturers exert influence on the process of automotive intelligence?
Chen Boyu pointed out that as automotive electronic and electrical architecture enters the domain control stage, SoC, as a core component of algorithms, plays a crucial supporting role. Although the absolute number of SoCs in vehicles will decrease, it will rapidly develop towards higher integration and higher computing power. Socionext has launched a new IP module, providing platform-based SoC solutions in the fields of Advanced Driver Assistance Systems (ADAS) and smart cockpits to help customers quickly iterate and mass-produce their chips.
New IP modules focus on two major areas
Currently, whether it's Tesla or numerous emerging domestic car manufacturers highlighting their L2 and above-level assisted driving products, they all heavily rely on automotive camera sensors, with the ISP processor being a crucial component. Chen Boyu stated that, leveraging its deep technical and experience accumulation in the consumer electronics field, Socionext has recently developed an automotive-grade image signal processing module called "Sensight." Sensight provides image preprocessing in smart cockpit camera applications, specifically including image denoising, restoration, recovery, and distortion correction; in ADAS assisted driving system applications, Sensight can filter invalid image information collected and captured by vehicle sensors, ultimately extracting valid information such as vehicles, pedestrians, lane lines, and road traffic signs, providing support for ADAS scenarios such as lane detection, 360° surround view systems, automatic parking, and driver fatigue warning systems.
In the field of visual sensors, many solution providers are attempting to overcome the technological bottlenecks of cameras through AI and computing power. Against this backdrop, Socionext has launched the "NXA Module," a deep learning AI accelerator designed to meet the current deployment needs of L1-L3 level assisted driving scenarios. Chen Boyu explained that the Socionext NXA AI accelerator module comprises two parts: DPA and NNA. On one hand, it uses convolutional neural networks to recognize and match people, animals, vehicles, road signs, and various other obstacles. On the other hand, it allows for flexible selection of NNA50, NNA100, and even NNA200 to adjust computing power and achieve high efficiency with low power consumption. Currently, the IP's energy efficiency is 12.1 TOPS/W, which is far ahead of competitors in the industry.
Adhering to the principle of leading the automotive ASIC field through technological advancement
Socionext has long been committed to the research and development of high-performance and high-quality products. The company has consistently kept pace with the evolution of advanced technologies, providing leading automotive-grade chips to customers in China and globally, playing a leading role in the industry. Chen Boyu explained that Socionext has already achieved mass production of automotive-grade chips on 28nm, 16nm, and 7nm process platforms, and has also established a deep collaboration with TSMC on its industry-leading 5nm automotive-grade chip. The product is currently in the planning stage, and samples are expected to be provided to the market in the second quarter of next year.
Unlike typical custom chips, automotive-grade custom chips place greater emphasis on automotive-grade quality control, certification, and processes. Socionext employs designs that prevent defects and reduce manufacturing flaws, and through close collaboration with manufacturing subcontractors and various quality management measures, is committed to providing high-quality, high-reliability products. Socionext offers a range of services including functional safety (ISO26262), testing, design verification (DFT/DFM), quality assurance, and analysis technologies, providing high-quality automotive chips for automotive electronic systems.
The improvement in the level of automotive intelligence has driven the upgrade of its EE architecture from the previous distributed ECU architecture to a centralized domain controller architecture, and continues to evolve towards a centrally integrated architecture.
In the centralized domain controller architecture stage, traditional MCUs cannot meet the complex electronic and electrical architecture and massive data processing requirements. SoCs have become the mainstream trend in automotive chip design and application due to their advantages such as improved computing power, high data transmission efficiency, reduced chip usage, and more flexible software upgrades.
Since 2021, the rapid growth in sales of new energy vehicles, coupled with the development of AI on the automotive side driven by autonomous driving, has increased the demand for computing chips, leading to a period of high growth for SoC chips. The global automotive-grade SoC market is expected to maintain a growth rate of over 35% from 2021 to 2025, with a projected CAGR of 27% from 2024 to 2028.
China has a certain leading advantage in automotive intelligence, with a relatively high demand for SoCs globally and a growth rate higher than the global average. The Chinese automotive SoC market size is expected to remain above 40% from 2021 to 2025, with a projected CAGR of 28% from 2024 to 2028.
Autonomous driving SoCs are SoCs specifically designed for autonomous driving. According to Black Sesame's prospectus, the CAGR of the global and Chinese ADAS application SoC market size was 40% and 55% respectively from 2019 to 2024, and is expected to be 24% and 23% respectively from 2024 to 2028.
Based on the AI computing power requirements of the main control SoC chip for different levels of intelligent driving solutions, intelligent driving SoCs can be divided into low computing power chips (2.5-20 TOPS), medium computing power chips (20-80 TOPS), and high computing power chips (≥100 TOPS). Low computing power chips mainly focus on basic L0-L2 assisted driving functions, and some models can provide NOA (Noise, Arrival, and Assist) functions. The price range of vehicles equipped with these chips is 100,000-150,000 yuan, and their characteristic is to pursue high cost performance.
The medium-power computing chip is mainly a lightweight integrated domain controller solution for driving and parking, characterized by high-speed NOA, city memory NOA and memory parking functions. Some models can provide city NOA, and the price range of the models equipped with it is 150,000-250,000 yuan.
High-performance computing chips support advanced integrated domain controller solutions for driving and parking, and even integrated cockpit solutions, enabling "easy-to-use" L2+ level functions such as city NOA and AVP. Some models consider pre-installing hardware to achieve L3 and higher levels of autonomous driving functions, with models equipped with these chips priced above 250,000 yuan.
The development towards advanced intelligent driving functions requires new algorithms (Transformer+BEV+OCC) and a more advanced vehicle EE architecture (central computing + regional control), all of which require more powerful SoC chips as the "foundation" for support.
