Recently, Ni Guangnan, an academician of the Chinese Academy of Engineering and a researcher at the Institute of Calculation Technology of the Chinese Academy of Sciences, once again caused heated discussions in the semiconductor industry. He said that the RISC-V can well meet the ideal choice of "demand definition software, software definition hardware" intelligent networking car chip is the ideal choice is that it is Time needs.

From the perspective of the industrial development pattern, the RISC-V automotive industry does have huge development opportunities. If it rationally guide its open source and modular characteristics, it can better adapt to the general trend of the development of the automotive industry. At the same time, we should also consider the factors of the automotive industry and the higher the threshold of automobile supervision chips.

In traditional cars, RISC-V can be roughly divided into three categories: functional chip MCU, sensor BPV11F, and power semiconductor. Among them, the functional chip MCU is mainly responsible for the information transfer and data processing of the car in the related systems such as the body's electronic system, the body power system, and the entertainment system.

With the improvement of smart cars, the scenes that cars need to face and the amount of data generated by the car have changed dramatically. Therefore, in addition to functional chips, car chips have changed from three types of traditional categories to the latest five types of MCUs, adding power semiconductors, sensor AI chips and memory.

On the other hand, the main application direction is MCU and AI chips.

At present, the market is already based on the RISC-V architecture MCU chip. If these chips want to win opportunities in the automotive field, the first step is to enter the industrial chain and enter the pre-installed market.

Of course, for RISC-V as a architecture, it does not necessarily appear in a single product in MCU. Under the overall trend of heterogeneous, it can also be used as a soft nuclear or hard core inside the chip to improve the rich IO interface richness of related chips. However, whether it is the core or as the core IP, it is also necessary to cross the car regulations.

Academician Ni Guangnan mentioned "software definition hardware" In fact, there is another saying in the industry: software definition cars. With the rise of the concept of intelligent networking cars, the application of software in the automotive industry is becoming more and more common and important.

The essence of software definitions is the large upgrade of automotive electronics and electrical structure EEA. From the traditional distributed ECU to the domain controller, and gradually transitioning to the statistical central centralized processing platform, this is why people say that in the future, cars will become more and more like a one The computer is driving on the road.

During the upgrade process, the electronic electrical structure EEA is also changing. From the past, the information processing chip on the car has also changed. From the past, the MCU computing power has become more and more powerful AI chips.

When the central centralized architecture is implemented, the car will enter the 3.0 era, autonomous driving, ADAS, smart cockpit and networking of the car will become the core function of the car, and all functions can be upgraded through OTA. Of course, car 3.0 is not the ultimate form. There is a car 4.0 in the back, that is, the concept of further strengthening the network of the car, and placing the software in the cloud to further improve the computing power level of the vehicle.

If you want to complete the conversion of automobile 3.0 and car 4.0, you need to abstract and decide the car hardware through software, which is the core nature of software defining cars. Of course, in this process, other auxiliary elements need to be added, such as middleware, operating system and virtualization.

Although it is called the central centralized architecture, this does not mean that the car AI chip used is only CPU, as well as GPU, FPGA, ASIC and adding neural network units N-SOC, etc. Like the MCU, these chips are not necessarily necessarily necessarily. The form of a single product is on the car. Under the concept of heterogeneous computing, the manufacturer is also seeking various integrations to improve the flexibility of hardware and better adapt to the new ecosystem of software continuously upgraded.

Therefore, the software defines the automotive era and automobile era. In addition to continuously pursuing large computing power AI, in order to cope with the explosive growth of non -structured data on the car, a necessary feature of the chip is flexibility.

Then, in order to deal with the complex needs of software, RISC-V has many outstanding advantages. RISC-V is a loose open structure BSD (Berkeley software kit) open source mode allows it to be used freely at any purpose, allowing anyone to design, manufacture and sell RISC-V chips or software. The RISC-V instruction set is very concise. There is no image ARM architecture and the same historical burden of the X and 86 architecture, which has strong scalability. The RISC-V has the advantages of modularity and can be organized in a modular manner to better meet the high flexibility requirements of software-defined hardware.

Intelligent networking cars are an important terminal market for future chips and need to be planned in advance.

Risc-V, there are many obvious difficulties that need to be leapfrog to the Risc-V to achieve achievements in intelligent networking cars before getting on the bus.

The first is the long -term certification of the vehicle specification level. Generally speaking, it usually takes about 2 years to complete the vehicle specification -level authentication for a chip. After entering the automotive company's supply chain, the 5-10 years of supply cycle RISC-V is characterized by structure-related companies, this industry is very unfriendly, but it is also necessary.

The second is the requirements of computing power. At the core level, the RISC-V is sprinting the high-performance computing market, but there is no scale yet, and there is still a gap between the ARM architecture. This is also an obstacle. The threshold.

The third problem is the coordination of the industry chain. We mentioned above that software -defined cars need intermediate parts. Let's take a look at Autosar, a representative middleware now, and the chip companies they participated in are basically Bingfei, Risa, NXP, ARM. At present, these companies are unwilling to let RISC-V get in the car, and Ruisa's RZ/FIVERISC-VMPU uses does not include the automotive field. In addition to the middleware, the RISC-V also needs to implement the interface standardization, adapting operating system, etc., and also requires the software and tools of related companies.

Conclusion

Deloitte predicts that it is expected that by 2025, the service market (SAM) of chip sales based on RISC-V technology will reach 2.9 billion in the automotive field (SAM). Therefore, the RISC-V on the car has a bright prospect, but Deloitte also predicts that the RISC-V is difficult to shake the status of the ARM and X86 architecture in the past 10 years. Then, in the automotive field, this cycle may be longer.

Of course, we cannot deny that Academician Ni Guangnan's point of view is wrong from the long cycle and difficulty. necessary.