Some people call RISC-V legendary, it was first proposed in 2010, more than ten years has been able to compete with ARM and X86, open source, free, lightweight is its characteristics. Nowadays, RISC-V has gradually built a good technical exchange ecology, and more and more "big players" are also deployed in it. Due to the chip blockade in the United States, the risk of using ARM architecture is increasing, so soon after the birth of RISC-V, it is also considered to be the key to China's "stuck neck" problem. Some have called RISC-V an investment scam. It belongs to the same reduced instruction set architecture as ARM, which is very similar in performance, but ARM architecture has accumulated more than 20 years of IP design experience, and has unparalleled advantages in consumer electronics and embedded. In addition, ARM also has a complete industrial chain ecology, which makes it difficult for latecomers to challenge. Therefore, there are many people who are not optimistic that RISC-V can eventually complete the replacement of the ARM architecture, and the prosperity of RISC-V is only "short-lived". Despite both encouragement and concern about RISC-V, RISC-V has taken shape after thirteen years of development. In July 2022, Calista Redmond, CEO of RISC-V International Foundation, announced at Embedded World that shipments of RISC-V architecture processors have exceeded 10 billion. Experts predict that in two or three years, RISC-V architectures will surpass existing ARM architectures and available performance. How did RISC-V go from 0 to 10 billion? In the instruction set "Battle Royale", can RISC-V make it to the finals and compete with ARM and X86? In this article, a number of industry experts are invited to analyze the development and change of RISC-V from technology to market, as well as the future prospects.
RISC-V moves from "humble" to perfect
RISC-V is an open instruction set architecture (ISA) designed from the work of computer scientists at the University of California, Berkeley. Unlike other ISAs, RISC-V is designed to be a general-purpose instruction set for a wide range of applications, including personal computers, servers, embedded systems, and supercomputers. At the same time, RISC-V also provides an open standard that allows different hardware and software to be compatible and interact with each other. Since RISC-V was first proposed in 2010, it has become a hot topic in the field of computer science, and more and more companies and organizations are adopting RISC-V for their computing needs. The key to RISC-V's belief that it can replace ARM lies in its open source and free strategy. However, compared with ARM or X86, RISC-V is still a relatively "young" instruction set architecture after all, and there will always be imperfections in the industrial chain and R&D tools. So why did manufacturers choose RISC-V, which is still in the "novice period"? Zhengyu Hu, General Manager of Codasip Greater China, said: "Although RISC-V is just getting started, many open source tools and software libraries are currently supporting RISC-V development, including compilers, emulators and other auxiliary tools, and there are many RISC-V specific development boards and system-on-chips on the market. As RISC-V technology continues to grow, its supporting tools will also be improved. Lin Zhiming, chairman of Andes Technology, also holds the same view, saying that in the RISC-V ecosystem, manufacturers are very willing to develop supporting tools. In the case of so-called deficiencies or imperfections, it should be short-term, and as RISC-V has a large global membership and all parties accelerate its development, the current imperfect or insufficient situation will actually gradually decrease over time. Also a streamlined instruction set, RISC-V wants to replace the mature ARM, it must make efforts on its own advantages. Looking through the RISC-V specification documents and tutorials, it is easy to find that it is quite different from the ARM instruction set in terms of simplicity, such as abandoning instructions such as "reading/writing multiple registers at once" and "conditional code", which can make it easier to increase the CPU main frequency. In terms of architectural complexity, RISC-V itself is also very simple, the basic instruction set is only more than 40, plus other modular extension instructions a total of dozens of instructions, its specification document is only 145 pages, and the "privileged architecture document" is only 91 pages. Compared with the cumbersome and complex ARM instruction set, RISC-V is indeed simple and easy to use, but in more complex CPU designs, will these features become a shortcoming? Hu Zhengyu said: "Although RISC-V's design philosophy is simple and efficient, it can improve programming efficiency, reduce power consumption, chip area and cost. However, this does not mean that RISC-V cannot meet the design requirements of complex CPUs, on the contrary, the company can customize and improve for various application scenarios, including fully optimized customization for complex applications or scenarios requiring high computing power, which is an advantage that general-purpose computing devices for the same scenario cannot have. We know that the ARM instruction set is operated and maintained by the ARM company itself, so its instructions are highly unified, and the chip factory can only make extensibility adjustments after applying ARM IP, and the chip design must also be done around the ARM core. RISC-V in this regard, on the contrary, open source can make each vendor highly customized to suit their own IP core, but the vendor's separate battle may bring new problems, that is, instruction set fragmentation. Fragmentation means that each vendor develops different instruction sets, and different products are incompatible with each other, which ultimately leads to ecological fragmentation. With more and more players already pouring into the RISC-V track, how can RISC-V prevent fragmentation? Lu Xiang, deputy general manager of Apt Microelectronics, believes that RISC-V does not need to worry too much about fragmentation. He pointed out: "CPU cores based on the RISC-V instruction set must have some differentiation due to the different needs of the target market. However, these differentiations can be measured by the performance and customer satisfaction of the final landing product to whether it is worth promoting or continuously evolving. Through the screening of the market, the final precipitation must focus on the implementation of the optimal solution, so there is no need to worry too much about fragmentation in the same market. At present, Apt's 32-bit MCUs based on RISC-V research and development and mass production have been widely used in industrial control, smart home appliances and other fields, and have been recognized by customers through excellent performance, so product performance is the core focus. RISC-V started late, but in a short period of time, RISC-V can develop to the extent that it is today, which is inseparable from the efforts of RISC-V International Association in standardization. In this regard, Lin Zhiming said that RISC-V's instruction standardization is very good, especially in terms of adding instructions, RISC-V is even much better than ARM. "RISC-V has grown rapidly over the past decade, and especially in the last two years, adding instruction sets at a rate of about thirty per year. RISC-V's update process is very standardized, from approval, announcement, and then to an open source open instruction set. These are actually being led by RISC-V International. Therefore, the RISC-V new directive can be normalized and published in a very short time. ”
Link: RISC-V moves from concept to application
In 2015, the Global RISC-V Foundation was established, announcing that RISC-V officially moved from concept to reality. In July 2018, the Shanghai Municipal Commission of Economy and Information Technology took the lead in promulgating relevant RISC-V support policies in the country, and since then the domestic industry has entered a period of rapid evolution. At that time, Guo Yiwu, secretary-general of the Shanghai Integrated Circuit Industry Association, said that RISC-V has the advantages of streamlining, low power consumption, modularity, scalability and other technologies, and in recent years, the RISC-V ecosystem is rising rapidly around the world, becoming an important innovation focus for the semiconductor industry and emerging applications such as the Internet of Things and edge computing. In fact, Secretary General Guo Yiwu must have some courage to make this judgment in 2018. At that time, the ARM architecture was in full swing, and after the company was acquired by SoftBank, in 2018, Son proposed to re-list ARM within five years; In terms of products, the Snapdragon 835 and 845 occupy almost all of the high-end mobile phone market; The number of chips using the ARM architecture also reached 22.5 billion that year. However, only 5 years later, the number of chips using the RISC-V instruction set has exceeded 10 billion, and there seems to be a trend of catching up with ARM. The number of RISC-V members has also reached more than 3,100, including Google, Huawei, NVIDIA, Qualcomm, Samsung, IBM and other global first-tier manufacturers are among the members. At present, from the Internet of Things to driverless cars, from embedded systems to supercomputers, RISC-V has begun to demonstrate its advantages in various application scenarios, but ARM still firmly occupies a monopoly on mobile terminals, how should RISC-V break through ARM's defenses? Lu Xiang said: "ARM's own advantage is ecology. ARM and Android have formed an A-A alliance, forming a large, complete and mature mobile ecosystem, which RISC-V cannot match. However, Google recently officially announced at the RISC-V Summit that Android will support the RISC-V instruction set. In other words, in the future, RISC-V chips can use the Android system. Once officially supported by Google, most Android applications will run on RISC-V chips without additional work from developers. Previously, considering the ecological problem, RISC-V could not replace ARM, now Android officially supports RISC-V, the ecology does not need to worry, it can be said that the best time to replace ARM has come. Apt is also working with Pingtou to build RISC-V ecosystem construction from CPU to SoC, chip design, tool chain, processor, software ecology and other chains. With the addition of more and more players, I believe it will have a wider range of application scenarios and development space in the future. Hu Zhengyu also looked forward to the future of RISC-V: "RISC-V will gradually expand to more application areas in the future, and the industry needs more secure and efficient processors to support complex requirements, and through the continuous development of RISC-V software ecosystem and series of tools to achieve continuous breakthroughs in technology." Lin Zhiming is also optimistic about the application of RISC-V, saying: "Some embedded systems of RISC-V have done better than ARM, and the next technical nodes to be overcome are Android, iOS or Windows, and these key technologies may take three to five years to make breakthroughs." ”
Challenge ARM, RISC-V manufacturers have clear ideas
Over the past few decades, ARM has become a mainstream instruction set in mobile devices, the Internet of Things, automotive, industrial automation, and more. However, as computer use cases continue to change and evolve, some problems are beginning to emerge, such as ARM's licensing fees and closed-source restrictions, which limit the application development and innovation of some companies and organizations. The advent of RISC-V offers a completely new solution. In just a dozen years, RISC-V has completed a major breakthrough from theory to practice, from niche to wide application. However, compared with ARM, RISC-V still has some deficiencies in ecology, community construction and upstream and downstream support. First of all, the ARM ecosystem is very mature, with a wide range of application scenarios and a large number of developers, so in some specific fields, especially in the mobile field, ARM still has a very big advantage. In addition, there are a large number of excellent software and tools in the ARM ecosystem, which makes it easier for developers to develop and optimize applications. Second, the development and improvement of the RISC-V ecosystem will take some time. While the RISC-V community is constantly evolving and innovating, its ecosystem is not mature enough relative to ARM and lacks some great software and tools. Therefore, in some specific scenarios, the application of RISC-V may be limited. Finally, the RISC-V ecosystem lacks some mature vendor support, such as ARM architecture licensors and chip manufacturers in the ARM ecosystem, which also makes RISC-V hardware development and production need to be further developed and improved. After the merger, how should RISC-V progress to break ARM's ecological blockade? Hu Zhengyu believes that the software ecology is very important for the improvement direction of RISC-V. The development of software ecology and hardware ecology complement each other, and the software ecology needs to improve even more, such as the common Linux kernel and GCC compiler. In addition, it may also be necessary to optimize virtualization and voice-related open source software, such as Docker, Kubernetes, network acceleration software stacks, AI software databases, etc. also need to do special optimization and performance tuning, requiring huge investment of manpower and material resources. In the process of RISC-V popularization, Chiplet technology has brought many development opportunities to China's integrated circuit industry. Chips are expensive when manufactured using advanced processes, and only a few foundries in the world can control high-end processes. In the case of limited production capacity and cost, mixing particles at different process nodes is a key development technology in the future, and Chiplet is also expected to become one of the magic weapons to break the blockade of foreign chips. Lu Xiang said: "If the integration of Chiplet technology and RISC-V can be realized, the entire industry will take a big step forward. However, this approach currently leads to each node becoming more complex and expensive, and if the industry wants to move to support integrated systems based on chiplets, different companies must start sharing chip IP with each other. Lin Zhiming believes that RISC-V should increase its investment in open source, business and education, he said: "First of all, follow the open source approach, which is the original UC Berkeley open source approach. After the relevant manufacturers accept such an open source ecosystem, the software and hardware developed by it must be able to feed back into the open source framework, so that the RISC-V ecosystem can be more advanced. The second is to conduct business operations from the perspective of reasonable profits. Finally, it is necessary to increase investment in RISC-V educational institutions or teaching and research institutions to expand the number of users. The three-pronged approach will greatly help the popularization of ecology. In general, RISC-V challenges ARM architecture is not a "mirror", RISC-V represents a new computer architecture idea and design concept, which will have a positive impact and promotion on the computer industry. We have reason to believe that with the continuous development and improvement of the RISC-V ecosystem, it will play an increasingly important role in the future computer field and become an instruction set architecture with wide influence and continuous innovation.