2023-01-03 《What Is The Main Control Chip Of The Automotive？》

The main control chip of the car is used to generate the calculation and generation functions of the main control signals of the car. The main control chip calculates the corresponding processing measures by receiving the signals collected by various sensors, and sends the driving signal to the corresponding control module. Therefore, the main control chip is equivalent to the "central" of the car chip group.

(1) Main control chip—MCU chip: the core component of the automotive electronic control unit ECU

MCU (Microcontroller Unit) is a microcontroller, also known as a single-chip microcomputer, which integrates the CPU, memory, I/O ports, serial ports, timers, interrupt systems, special function registers, etc. On, apply it in different products, so as to realize the calculation and control of the product. The on-board MCU is the core component of the automotive electronic control unit (ECU), responsible for the calculation and processing of various information, and is mainly used for body control, driving control, infotainment and driver assistance systems.

With the acceleration of automotive electronics and intelligence, the number of automotive ECUs is also increasing rapidly, driving the growth of MCU market demand. The computing power continues to increase with the improvement of intelligence, from L1 (automatic driving level) <1TOPS computing power to L5 (automatic driving level) 1000+TOPS computing power to promote the rapid growth of the main control chip. All ECUs need MCU chips. Usually, one ECU in a car is responsible for a single function and is equipped with one MCU. There will also be a situation where one ECU is equipped with two MCUs. Compared with traditional fuel vehicles, new energy vehicles have more functions. The demand for MCU performance, power consumption, and quantity has all increased. Traditional cars and bicycles will use about 70 on average, while new energy vehicles will need to use more than 300.

(2) Main control chip—SoC: a system-on-a-chip integrating AI accelerators, which is applied to automotive smart cockpits and automatic driving

The full name of the SoC chip is called: System-on-a-Chip, which means "put the system on one chip" in Chinese. Automotive electronic functions rely on on-board chips to achieve. With the implementation of ADAS and the maturity of L3 and above-level autonomous driving, traditional central computing CPUs cannot meet the computing power requirements of smart cars. A system-on-chip (SoC) that combines heterogeneous integration and AI accelerators has emerged as the times require, and is mainly divided into smart cockpit chips and autonomous driving chips.

According to IHS data, it is estimated that the global automotive SoC market will reach US$8.2 billion in 2025, and autonomous driving above the L3 level is expected to enter the market on a large scale after 2025, and supporting high computing power and high-performance SoC chips will bring extremely high additional value. It is expected to drive the rapid expansion of the main control chip market.

(3) Smart cockpit chip: a strong support for the future "one chip with multiple screens" technology

According to the analysis report, the global smart cockpit market is expected to reach US$43.8 billion in 2022. As the smart cockpit gradually evolves from the electronic cockpit to the third living space, the "one core, multiple screens" cockpit solution has become a future trend. Complying with the evolution trend of smart cockpit multi-sensor fusion, multi-mode interaction and multi-scenario mode development, as a processing The central cockpit SoC needs continuous development and breakthrough. Therefore, the computing power of the cockpit SoC will continue to increase. According to IHS Markit, it is estimated that the computing power demand of the cockpit NPU in 2024 will be ten times that of 2021, and the computing power demand of the CPU will be 3.5 times that of 2022. At the same time, the chip itself will also develop in the direction of miniaturization, integration, and high performance.

Now that software-defined cars, especially cockpit systems, are increasingly open, mobile applications are migrating to the cockpit in large quantities, so the number of operating systems on a single chip will be 3 to 5 times that of before. At the same time, because in the past, a central control system may only run one system and support one screen, but now the cockpit needs to support multiple operating systems and multiple screens. The advantage is that the entire application program will be 50-100 times that of traditional cars, but it also leads The complexity of the entire software-defined car has increased geometrically. Future intelligent technology cockpit scenarios include traditional central control entertainment navigation, LCD instrument panel, driving recorder, independent rear entertainment, head-up display, 360-degree surround view + automatic parking assistance, voice assistant is also essential, and it is also There will be driver monitoring systems as well as passenger monitoring systems, virtual air conditioning panels and more.

(4) Automated driving chip: the basis of computing power determines the level of autonomous driving

On the one hand, the self-driving chip needs to meet a higher safety level. On the other hand, with the improvement of the level of self-driving, the requirements for the computing power of the self-driving chip are also increasing. Chips with only CPU processors are difficult to meet the needs of computing power, and autonomous driving chips will develop in the direction of a heterogeneous SoC integrating CPU+XPU. XPU includes GPU/FPGA/ASIC, etc.

GPU, FPGA, and ASIC each have their own strengths. GPU is suitable for computing and processing data-intensive applications. FPGA realizes logic editability through redundant transistors and connections, while ASIC is a dedicated chip designed for specific user algorithm requirements. As a kind of ASIC, NPU simulates neurons at the circuit layer, and realizes the integration of storage and computing through synaptic weights. One instruction can complete the functions of hundreds of instructions on the GPU, improving operating efficiency. NPU has been widely adopted by many manufacturers. If the automatic driving algorithm is unified in the future, ASIC/NPU is expected to become the most efficient automatic driving chip solution.