On June 30, 2021, the China Automotive Semiconductor Industry Conference continued in Shanghai. Shu Jie, the Senior Manager of ARM China’s Automotive Market, shared and exchanged ideas on “The Computing Architecture and Technology of Software-defined Cars”. The following is a summary of the key information.
About ARM
ARM does not make chips. We sell IP to customers who then make the chips. Buying IP to make chips is a common business model in the semiconductor industry.
ARM takes on some of the R&D work of chip companies and helps with the design of CPUs, GPUs, and SPUs. Customers then integrate these designs into their own chips for production. ARM’s chip designs can meet the different needs of customers and provide great flexibility from an IP perspective.
In addition to IP products, ARM maintains a large ecosystem. I am involved in the automotive ecosystem maintenance work to help customers benefit from it and help them create great products and applications.
Next, I’ll provide some data to help you better understand ARM.
Currently, we have nearly 2,000 licenses worldwide and over 530 customers. The number of chips shipped based on ARM technology has exceeded 180 billion. In the past year, 23.7 billion chips based on ARM technology have been shipped. These figures are all from our customers.
As I mentioned, ARM itself does not make chips. We only authorize IP to our customers, who in turn incorporate the IP into their designs and then manufacture and sell the chips.
In the automotive ecosystem, ARM has been active since 1994 and many of our IPs can meet the needs of intelligent connected vehicles, such as entertainment information (IVI) systems, powertrain, ADAS sensors, and applications for vehicle body and chassis.
In IVI and ADAS, more than 60% of SOC applications’ computing cores are based on ARM technology. There are also some new applications, like the emergence of domain architectures. A few days ago, ST released a chip that uses R52 for domain controllers in chassis and power applications. It is in mass production and there were some sample chips before. ADAS chips also use ARM technology and computing cores. Currently, the top 15 automotive chip manufacturers are our clients.
ARM has been cultivating its new customers. In 2021, on average, there are about 18 ARM processors in each new car. ARM also focuses on public safety, upgrading computing power, and collaboration, which can help the automotive industry with digital transformation. At the same time, ARM also provides some information security solutions.
Software-defined SystemsWhen discussing software-defined systems, the first thing that comes to mind is the need for decoupling. Software needs to be independent of a specific chip, operating system, or vehicle model, and not be affected by hardware. This definition is based on the use of service-oriented architecture (SOA) to develop systems. Services can be delivered, published, and subscribed to independently, and can also be managed.
There are two development modes for using this software-defined system. The first is to extract software services from a single hardware platform by using technologies such as virtualization and containerization. The second is to abstract software services from a distributed hardware platform using a real-time bus. We have seen some design patterns for intelligent software economic systems.
Now let’s take a look at some of the current applications of software-defined systems. The main applications are currently in data centers, where there is software-defined networking and storage. In the future, the entire data center will be supported by technologies such as adapters, containerization, and virtualization, as well as enterprise-level operating systems. These are all applications of software-defined data centers.
This software-defined technology will extend to other applications, such as industrial automation systems for automotive software. There will also be some technologies to support autonomous driving applications, which will have over 300 million lines of code. Behind all this software, there is always the support of electronic and electrical architecture.
Application of Domain Architecture
In terms of domain architecture applications, there is first the cabin, and then there are some body controllers in use. For driving and other applications that are relatively slow, there will gradually be new products coming out. For future development, our cars will move towards a centralized architecture with an embedded high-performance computer.
In addition, through technologies such as Ethernet and edge modules, sensors and actuators can be connected to achieve control of the vehicle. In terms of trends, everyone will mention what kind of computing platform will be used, which is the most important concern, because there is currently no computing platform that is truly suitable or has been mass-produced for this specialized architecture.
So how is a software-defined system achieved? We will implement it from three aspects. First, we will focus on the ecosystem and information security, such as the platform standards or certification projects. Then there is the PSA project, which was originally developed for IoT and has now been expanded to vehicles. IoT is more complicated, so we will provide some best security practices and guidelines based on information security. This is the purpose of our PSA certification.
Security Lifecycle PlanningLet me talk about our security cycle plan. We have been carrying out this project for many years, with the goal of reducing customer design investment, accelerating deployment, and quickly producing the chip. If someone in China wants to make a certified chip for use in cars that meets public safety requirements, it will take a very long time. However, if we have the support of this type of contract manufacturing IP, certification will be much easier.
Source: 2021 China Automotive Semiconductor Industry Conference
This article is a translation by ChatGPT of a Chinese report from 42HOW. If you have any questions about it, please email bd@42how.com.