When it comes to x86 and ARM, many people might not be unfamiliar. Intel and AMD, based on x86, dominate the PC chip market, while ARM architecture has also begun to shine in mobile chips like Kirin and Snapdragon.
Compared to these two “top players,” another architecture that is favored by Chinese semiconductor manufacturers, RISC-V, may not be as widely recognized among the general public. However, in the eyes of industry insiders, it is seen as a “potential stock” with high expectations.
For example, Mr. Ni Guangnan, an academician of the Chinese Academy of Sciences, once compared x86, ARM, and RISC-V, believing that RISC-V is likely to develop into one of the world’s mainstream CPUs, thus forming a pattern of three powers in the CPU field: Intel, ARM, and RISC-V.
What gives this architecture, which has low public recognition, the courage to compete with giants? And why does a technology ecosystem from the United States have such deep ties with China?
Trying to compete with x86 and ARM:
What exactly is RISC-V?
To be honest, although RISC-V has been popular for many years, it might still be a mystery even to industry insiders what it really is.
In 2020, Krste Asanovic, chairman of the RISC-V Foundation, had to personally write an article titled “RISC-V is not an ‘open-source processor’” to clarify the definition of RISC-V, which is quite unfortunate for this architecture.
So, what exactly is RISC-V? Let’s try to clarify it in simple terms:
First, the official definition is clear: RISC-V is an open instruction set architecture based on the principles of “Reduced Instruction Set Computer (RISC).”
It’s important to note that RISC-V is not a processor or chip (Implementation), but rather an instruction set specification (Specification).
The instruction set is the interface between software and hardware, guiding the CPU on how to perform calculations. As a standard specification, how it is used to design chip-level software systems to help CPUs run more efficiently is up to the manufacturers’ capabilities.
(Analysis of the relationship between instruction sets, microarchitecture design, and products)
Secondly, RISC-V’s rise can be attributed to its inherent characteristic of being “born for open source.”
There can be many types of instruction sets supported by a CPU, and before RISC-V appeared, there were already various instruction set architectures (ISA), whether it was the currently dominant x86, ARM, or DEC, IBM 360, MIPS, SPARC, etc. How did RISC-V, as a “transfer student,” manage to overtake them? Thanks to its completely open nature.
In 2010, a research team at Berkeley, when choosing an instruction set for a new project, found that many existing instruction sets had intellectual property restrictions. x86 was closed off by Intel, and ARM charged high licensing fees, so they decided to design a brand new instruction set from scratch.
In just three months, the four members completed the development of RISC-V and decided to open it under the BSD (Berkeley Software Distribution) open-source license. This is a highly permissive license, allowing users almost no restrictions; anyone can design their own processors based on RISC-V without having to pay licensing fees. This makes RISC-V particularly unique.
Moreover, it was developed under the concept of a reduced instruction set, which allows for lower CPU complexity and enables the production of more powerful CPUs at the same process level. Naturally, it quickly accumulated a large number of open-source practices and chip tape-out cases.
On this basis, RISC-V began to gain its most important asset—community ecology.
Every type of chip needs the support of an ecosystem to achieve commercial scalability. For example, the strength of x86 comes from Intel’s years of nurturing the server chip ecosystem. The instruction set serves as a bridge between software and hardware, and as an “intermediary,” it naturally requires ecological forces from both ends.
With the technical maturity of the RISC-V instruction set, the University of California, Berkeley, established the non-profit RISC-V Foundation in 2015 to build the RISC-V ecosystem. Unsurprisingly, the membership has been growing at an annual rate of over 100%.
Google, Qualcomm, IBM, NVIDIA, NXP, Western Digital, Microsemi, the Computing Institute of the Chinese Academy of Sciences, MIT, the University of Washington, UK Aerospace Systems, Huawei, Tesla, Samsung, Hitachi, Seagate, Alibaba, MediaTek… well-known enterprises and research institutions have joined.
With semiconductor design companies, system integrators, equipment manufacturers, military enterprises, and research institutions all contributing to the RISC-V ecosystem, toolchains and applications are naturally flourishing.
Last year, Tesla considered using free RISC-V designs in its new chips; NVIDIA is also planning to use RISC-V for internal GPU controllers; Samsung has disclosed plans to adopt SiFive’s RISC-V cores in various chips for its flagship 5G smartphones.
The RISC-V ecosystem is gradually growing and improving, and it can even compete with ARM in certain niche markets. Thus, it may not be surprising for people to see it becoming one of the “three major CPUs.”
Meeting at last: The Chinese connection of RISC-V
As an important open-source architecture, RISC-V has naturally received heavy support from the government eager for semiconductor industry advancement.
For instance, processor-related projects funded by the Indian government are starting to align with RISC-V, making it a de facto national instruction set; the Pakistani government has also announced that RISC-V is a preferred national architecture. However, when it comes to the country most closely tied to RISC-V, China certainly ranks high.
On one hand, this is reflected in the degree of industrial integration.
More than 20 domestic enterprises and institutions, including the Computing Institute of the Chinese Academy of Sciences, Huawei, and Alibaba Group, have joined the RISC-V Foundation. Berkeley and Tsinghua universities have also established the RIOS Lab in Shenzhen to support the RISC-V software ecosystem. In July 2018, the Shanghai Economic and Information Commission issued the first domestic policy supporting RISC-V.
Data shows that over 300 companies in China are paying attention to or developing based on the RISC-V instruction set. To date, many companies have established key technologies for open-source chips based on RISC-V and launched related products, such as the Xuantie 910 processor IP core launched by Pingtouge Semiconductor.
On the other hand, this is reflected in the dynamics of international public opinion.
For example, during the trade war, whether RISC-V processors were subject to U.S. export control regulations sparked considerable debate. Although RISC-V processors independently developed by companies outside the U.S. (including Europe and China) are not subject to substantial export control restrictions, the RISC-V Foundation still decided to move its headquarters to Switzerland, which clearly sets it apart from other American tech companies.
So, beyond the open and free philosophy, what kind of chemical reaction has occurred between China and RISC-V is something that needs to be logically traced.
Currently, it seems that the core reasons for the continuous connection between RISC-V and China can be summarized in three aspects:
The first reason is RISC-V’s unique advantages.
Open-source instruction sets are not uncommon; why has RISC-V gained such attention? The commonality of being free and open is certainly not enough. Compared to previous instruction set architectures, RISC-V has three particularly distinctive features: First, it is simple. x86 and ARM, as commercial architectures, have to retain many outdated definitions to maintain backward compatibility. In contrast, RISC-V, as a newcomer, does not have these historical burdens, and thus its instruction set documentation is relatively shorter, allowing developers to get up to speed more quickly.
Secondly, RISC-V’s modular architecture provides manufacturers with greater flexibility and the premise for customized production. RISC-V is the first architecture designed to allow users to choose appropriate instruction sets based on specific scenarios; different parts can be organized modularly, like a platter that allows for selective needs, thus fulfilling various application demands with a unified architecture. This scalability can reduce chip development cycles and thresholds, allowing smaller companies to participate, thereby enhancing manufacturers’ competitive edge.
In contrast, the Application, Real-Time, and Embedded categories in ARM architecture are not compatible with each other and cannot be custom-designed.
The second reason is the strong embrace of Chinese manufacturers.
The success of RISC-V in China is closely related to the current situation where our chip development is constrained. Based on open-source RISC-V, creating chips with independent intellectual property rights and fostering the corresponding industrial ecology is undoubtedly a refreshing wave in the desert for China’s semiconductor industry.
Especially in an unclear political environment, where over 90% of the global server chip market is built on Intel’s x86 architecture, and ARM has been acquired by Japan’s SoftBank, although IP licensing can be purchased, the core capability to design CPUs is lacking, making the sustainability of an industry based on licensing uncertain. Since the RISC-V instruction set itself is not a commodity, it naturally faces less controversy.
In addition to policy risks, the booming artificial intelligence construction in China also drives domestic manufacturers to embrace RISC-V. For example, AI hardware, such as smart speakers and smart home devices, requires greater edge computing capabilities, which demands more flexibility in architecture, necessitating constant adjustments based on market and technological changes. RISC-V can meet the demand for continuously adding new instructions, while architectures like ARM find it challenging.
With these factors combined, it is not surprising that RISC-V has become a “good match” in the eyes of Chinese manufacturers.
The third reason is the fertile ground of the Internet of Things.
If the previous reasons explain why China embraces RISC-V, then RISC-V also has a deep dependence on this land.
We know that the x86 instruction set basically dominates the PC market, while the ARM instruction set occupies most of the mobile processor market. In the territory of others, it would be difficult for RISC-V to shake up the market. The emergence of the AIoT field provides RISC-V with unprecedented opportunities.
On one hand, neither X86 nor ARM has a first-mover advantage in this field, let alone dominate it, which provides RISC-V with a premise to compete. Moreover, the requirements for the software and hardware ecosystem in the Internet of Things are not as high as for smartphones, and the industry chain is short and vertically oriented; the fragmented and flexibly configurable characteristics of RISC-V, similar to Android, also determine that it is more suitable for free flight in this domain.
In addition, IoT manufacturers are more cost-sensitive, and RISC-V eliminates expensive instruction set licensing fees, activating more companies that do not have the ability to design SoCs to join the ecosystem and develop diverse RISC-V devices, thereby invigorating the entire development ecosystem through sufficient innovation.
In fact, RISC-V architecture chips on the market, such as Huami Technology’s AI chip Huangshan No. 1 and Zhongtian Microelectronics’ CK902, are all targeting the vast IoT market. Clearly, when ecosystem construction is driven by a country and the entire industry, this momentum is much stronger than that of any single giant. Thus, it is only logical for RISC-V to focus on China for rapid development.
If the x86 era was about the WinTel battle for PCs and servers, and ARM + Android fought for the smartphone market, then RISC-V’s embrace of the Chinese industry will be bonded by AIoT for a long time.
The next adventure island:
The long-term outlook of RISC-V and China’s semiconductor industry
Behind the close bond, how to walk hand in hand towards the future is of great significance for China’s related industries.
At this point, we may need to temporarily detach from the emotions of “being in love” and consider some “settling down” issues.
First of all, the RISC-V ecosystem is still not as complete as ARM and x86, and is currently mainly applied to relatively low-end products, with no widely applicable cases in high-performance server CPUs or GPUs.
This is partly due to the need for corresponding software and toolchains based on RISC-V chips to be improved. Open-source instruction sets are not plug-and-play; just because the instruction set is open-source does not mean the CPU core is simultaneously licensed, which poses significant design and research requirements for chip companies. Clearly, in the face of the tough challenge of chip design, RISC-V has only helped the Chinese industry take a small shortcut, while a long and arduous path still lies ahead.
Secondly, under the premise that forward-looking technologies have not yet been conquered, RISC-V has already shown a tendency for excessive marketing in China. The old tradition of “once it’s open-source abroad, it’s autonomous domestically” has been fully fermented, and in the anxiety of being “choked” by chips, excessive hype and the establishment of joint ventures have been used to claim autonomy and control. For instance, the rampant promotion of “RISC-V is open-source” in China has blurred the lines; the chairman of the RISC-V Foundation specifically wrote an article clarifying this, indicating that these confusions could not remain silent any longer.
This kind of excessive hype that consumes public trust may undermine the credibility of RISC-V and related products, thus hindering the advancement of Chinese chips.
Furthermore, the usability of any architecture in the industry requires the support of a vibrant and diverse innovation in the industrial chain. While RISC-V grants manufacturers the freedom to design hardware, it may also lead to fragmentation issues similar to the Android system, where the diversity of devices and lack of backward compatibility can result in inconsistent standards and a fragmented development ecosystem.
This could lead to situations where RISC-V architecture processors from different chip manufacturers are unable to adapt to the same software during practical applications. Although this serious fragmentation issue has not yet occurred, the characteristics of AIoT networks make this phenomenon almost inevitable.
A strong leading manufacturer, incubating developers in a vertical ecosystem, standardizing the majority of code while leaving room for core application expansion, and establishing related application markets and basic platforms may drive the formation of a RISC-V innovation ecosystem more quickly.
Finally, the security issues of RISC-V’s core application scenarios in IoT must begin to deliver complete industrial solutions.
The security mechanisms of SoC chips are often established by hardware-enforced isolation of programs, data, and storage, creating a trusted execution environment; along with unique certificates and keys as the root of trust, plus secure boot and a series of tools to ensure system security. The RISC-V system instructions and privileged instructions lack security instructions, relying on software-defined domains MultiZone to ensure stable operation through hardware reinforcement.
In this context, for RISC-V to thrive amidst the AIoT and 5G wave and compete with ARM and x86, it must excel in security to gain the trust of chip manufacturers and ordinary users.
Recently, the RISC-V Foundation established the RISC-V Security Executive Committee to discuss security improvement solutions for IoT devices, embedded systems, and machine learning based on the RISC-V instruction set. How China will respond remains to be seen.
(Growth trend of RISC-V CPUs and major application markets)
Analysis firm Semico Research pointed out in its report “RISC-V Market Analysis: Emerging Markets” that by 2025, the market is expected to consume a total of 62.4 billion RISC-V CPU cores. The average compound annual growth rate of RISC-V CPU cores from 2018 to 2025 is projected to reach 146.2%. Among them, the 5G mobile phone, communications, and industrial sectors will bring new opportunities for companies investing in RISC-V.
Today, our discussion of RISC-V’s popularity in the Chinese market also proves that Chinese semiconductor companies have reached a new level, possessing the confidence and potential to push towards a higher industrial ceiling. Besides RISC-V, architectures like Loongson and MIPS are also on the list of alternatives for industrial independence.
This top-level architecture competition will undoubtedly unfold many prosperous details worth our continued attention.
