With X86 and ARM, why is RISC-V still popular in automobiles?

There is a joke in the software circle: Do you dare to touch ten-year-old code?

The specific situation of this story is: when a new colleague is told to maintain an old product, looking at the code package is like looking at flowers in the fog. When he goes to ask colleagues with more seniority, he finds that after several reincarnations, no one understands the specific logic, and he doesn't know how many companies the original author has jumped to. So he has no choice but to package it on the outside and deliver new functions.

This is the burden of history.

The advantage of RISC-V is that, as a rising star, it is flexible, streamlined and has lower development costs.

Today's cars , as "computers on wheels", need to include more than just passive safety. Infotainment has become a new requirement for automobile manufacturing, and software-defined cars are its new direction.

By comparing the application differences between X86 , ARM and RISC-V , we can clearly understand why RISC-V has taken the lead in the automotive sector.

01  X86 , ARM , RISC-V indicator comparison

From the power consumption and application of the three, we can find that X86 is more suitable for high-performance computing on the PC side, with the highest power consumption; ARM is suitable for the field of mobile communications, with less scalability than X86 but moderate energy consumption; RISC-V is mostly used in smart wearable devices, with streamlined instructions and no historical baggage, which means it has the lowest energy consumption. We know that cars have no power supply, but many devices such as the electronic control system and sensors on the car are power tigers, and intelligent driving will still consume a lot of electricity, so the low power consumption of RISC-V is a highlight. RISC-V avoids the detour of computer architecture development for decades, and the architecture is short and powerful, and the modules can be compatible, making the development cost lower.

The advantages of RISC-V open source are well known. On the one hand, it can flexibly expand the instruction set, which is very consistent with the fragmented and customizable characteristics of autonomous driving scenarios. The intellectual property rights of x86, ARM and other architectures belong to a specific foreign company, and it is impossible to modify and expand the instruction set architecture according to one's own ideas. On the other hand, because overseas companies such as ARM, Synopsys, and Cadence occupy the vast majority of IP patents, open source RISC-V will not be subject to the risk of IP supply interruption. The newness of autonomous driving and the newness of RISC-V just fit in.

Another major difficulty in the automotive market is the various stringent certifications. Now that RISC-V's security IP* has passed the automotive certification, its RISC-V core can already meet the high-performance requirements for automotive use. For example: Rambus's hardware root of trust IP RT-645 has passed ASIL-D certification; Andes Technology's currently launched automotive functional safety IP core - N25F-SE, is also about to pass ISO 26262 ASIL-B certification.

Now with the support of RISC-V, the pace of software-defined cars is accelerating.

02 Application of RISC-V in Automobile

The control chips of automotive chips mainly include MCU and AI chips. AI chips usually require higher processing performance, while edge applications require low-power, small-size MCUs. RISC-V happens to be an ISA suitable for general-purpose MCUs and dedicated accelerators for specific fields (such as AI).

MCU chip

As the core of embedded systems and motion controllers, MCU provides an application control layer, executes bus communication protocols, and provides a user interface. However, the development of MCU chips requires IP licensing. China has been trying to develop independent and controllable MCUs for many years, but it is still subject to others in terms of architecture. It should be noted that ARM licensing fees are high. Arm's license fee can account for 2%-15% of the chip price. Based on a 50% gross profit margin, it can account for 4%-30% of the chip cost. The emergence of RISC-V has ignited new hope for domestic MCUs.

MCUs designed with the RISC-V instruction set allow chip manufacturers/developers to quickly complete low-threshold, low-cost chip design, and can customize instruction design for specific application scenarios, with strong flexibility. In addition, RISC-V is also considered to be a good platform for long product life. It can be said that RISC-V and MCU are a perfect combination.

AI Chips

In the field of smart cars, the application of AI chips has been developed for some time. "The higher the computing power, the smarter the car" has become the slogan of many car companies.

In many scenarios such as human-computer interaction, visual processing, and intelligent decision-making, cars need to process large amounts of unstructured data such as pictures and videos, which places higher demands on on-board computing power. At this time, some people may compare RISC-V with X86. RISC-V is indeed streamlined, but some people may also raise questions: it does not have the stable high-performance computing capabilities of X86. How can it stand out in the face of the needs of smart cars?

Starting from reality, take autonomous driving as an example: Today's autonomous driving is still mainly concentrated in the L2 level. The computing power of various companies and new cars is not the current level of "autonomous driving assistance", but the level of "autonomous driving assistance" in the next few years. Using high-computing chips is equivalent to pre-burying a kind of ability in the car, which can be continuously unlocked in the future. A high-performance chip is not a rigid demand now, but a rigid demand in the future. You know, the development towards high performance is also an inevitable trend in the development of RISC-V. What's interesting is that RISC-V and smart cars belong to the same new track, and AI chips and smart cars are also promoting each other.

03 Constraints of RISC-V in Automotive Applications

The lack of a software ecosystem and the automotive industry’s “nostalgia” have increased resistance to the introduction of RISC-V into vehicles.

Although we are talking about software-defined cars, software support is precisely the shortcoming of RISC-V. Without a software ecosystem, chips based on RISC-V are useless in any field. In addition, the automotive industry has very high requirements for safety, which also makes the automotive industry less fond of new things. In addition to the disadvantages of the software ecosystem, related compilers, development tools and other ecological elements are still under development, making it more difficult for RISC-V to be installed in cars than for domestic manufacturers to use ARM.

RISC-V related companies cannot always want to free ride, but not actively invest in the construction of the ecosystem. Well, the same is true in the automotive field. At this stage when RISC-V is just emerging, it is particularly important to build the upper-level software suitable for RISC-V and how to make the flexibility of RISC-V fully unleashed in the automotive field, given the issues of concept, investment and commercial benefits.

Regarding the rapid development of smart cars, Academician Ni Guangnan of the Chinese Academy of Engineering also said that RISC-V can well meet the needs of the times of "demand defines software, software defines hardware" and is an ideal choice for smart car chip design.

04 RISC-V is already on board

In April 2020, Renesas Electronics Corporation, a major supplier of semiconductor solutions, and SiFive reached a collaboration to jointly develop next-generation high-end RISC-V solutions for automotive applications.

Subsequently, at the end of 2020, Renesas launched the automotive MCU RH850/U2B with the help of Japanese semiconductor company NSITEXE, which integrates RISC-V coprocessor. It is mainly used in hybrid ICE and xEV traction inverters, high-end regional control, connection gateways, vehicle movement and other related applications.

Earlier this year, Intel launched the Eye Q Ultra system-on-chip designed specifically for autonomous driving at CES 2022, but it should be noted that the Mobileye EyeQ Ultra does not contain any x86 cores, but instead has 12 RISC-V cores, Arm GPU and DSP.

Recently, SiFive announced the launch of three automotive-grade cores: E6-A, X280-A and S7-A solutions to meet the key needs of current and future applications such as infotainment, cockpit, connectivity, ADAS and electrification.

05 RISC-V and China's favorable time, location and people

Time

Under the current semiconductor development pattern, the ecological barriers of ARM and X86 architectures are high and solid, and the market share has been divided up by more than 80%. China has been tightly stuck in the original development path. Therefore, when RISC-V appeared, it became popular all of a sudden. RISC-V is suitable for the development of the industry and is more suitable for the Chinese market. At a time when 5G, Internet of Things, automobiles and other applications are blooming in many places, the emergence of RISC-V is the time for China to shine.

Geographical advantage

We should attach great importance to the RISC-V instruction set. The country should continue to strengthen policy support from the overall layout and top-level design, and implement and promote it step by step and in stages. As early as July 2018, the Shanghai Economic and Information Commission issued the first domestic policy to support RISC-V. In September 2020, China opened the Atom Open Source Foundation and disclosed its organizational structure, mission, vision and values ​​to the public for the first time.

Today, more than 20 domestic enterprises and institutions, including the Institute of Computing Technology of the Chinese Academy of Sciences, Huawei, and Alibaba, have joined the RISC-V Foundation. Berkeley and Tsinghua University have also established the RIOS Laboratory in Shenzhen to support the RISC-V software ecosystem.