October 15, 2022 Source:Leiphone View:368 Author: Wu You

BUCKING THE TECHNOLOGY INVESTMENT TREND 

In the seven years since RISC-V was introduced to China, many CPU enthusiasts have eagerly migrated to its open source and open features. 

Some saw the possibility of developing Chinese-specific IP cores based on RISC-V and promoting the commercial implementation of RISC-V through the sale of those IP cores to the China market. 

Others saw an even bigger opportunity. By stepping directly into the global high-performance CPU processor market with their own RISC-V chips, certain developers understood the advantage of bringing value to the RISC-V ecosystem and helping to turbo-charge the existing open-source Linux operating system. 

SpacemiT is one of those companies.  

Established in 2021, SpacemiT chose to focus on the development of a new generation of RISC-V high-performance processors and computing systems with more integrated architecture, more powerful computing power and better performance. 

Bucking the current downward trend in technology investment capital, SpacemiT has managed to obtain hundreds of millions of yuan in investment and financing. They have completed a Pre-A+ round of financing led by Legend Capital, along with by Matrix Partners, Brizan Ventures and Beyond Ventures. This is in addition to a previous Pre-A round jointly led by Matrix Partners and Glory Ventures. The list of SpacemiT shareholders also includes Zoo Capital, V Fund, Zhen Fund and several other well-known VC institutions.  

A team at the leading edge 

When compared to the big chip startups that have been favored by investment capital over the past two years, the two founders of SpacemiT are seen as “special” chip entrepreneurs.  

With no traditional overseas study or direct international experience working in a chip company, both Chen Zhijian and Sun Yanbang have found their backgrounds have been no obstacle to success. 

Dr. Chen Zhijian graduated from Zhejiang University and entered the semiconductor industry under the guidance of his mentor Yan Xiaolang. He has been engaged in research and development and the industrialization of domestic CPUs since 2006. With more than 10 patents and more than 20 papers in the CPU field, he participated in the research and development of C-Sky Microsystems. With the acquisition of C-Sky Microsystems by Alibaba, Dr. Chen joined Alibaba T-Head managing the R&D team as well as the application team for the domestic embedded CPU Xuantie series. 

His partner Sun Yanbang graduated from the University of Electronic Science And Technology Of China and was engaged in chip software development and customer ecosystem work for Allwinner Technology. With 13 years at Allwinner he gained a deep experience in software research and development. Sun has more than 10 patents in the field of basic software and has witnessed the successful mass production of perhaps hundreds of millions of chips. 

Of particular note has been the support they received from giants in the semiconductor industry such as Gao Bingqiang and Tang Lihua.  

Tang Lihua was one of China’s earliest chip entrepreneurs. He successively participated in the founding of Actions and Allwinner Technology and led the company as they went public. 

Gao Bingqiang is a world-renowned semiconductor chip expert. He was a Professor and Dean at the University of California, Berkeley and at Hong Kong University of Science and Technology. He has successfully invested in Da-Jiang Innovations, Montage Technology, SmartSens Technology, Broadcom Corporation, Narwal and many other semiconductor and product companies.  

In a recent interview with Leiphone, Sun Yanbang said, “The support of these semiconductor experts is mainly due to our team’s practical experience from R&D to mass production of RISC-V cores and chips, which is a scarce factor in the current Chinese semiconductor startup field.” 

This practical experience meant the RISC-V chip D1 developed by Allwinner Technology and based on T-Head Xuantie 906, could be mass-produced in early 2021 allowing the D1-based AIoT development board Nezha to become the world’s first Linux mass-produced product that supports 64bit RISC-V.  

“This is only a small step for SoC, but a big step for RISC-V,” added Sun Yanbang. 

Compared with other development boards, D1 provides a large-scale mass production supply and is easy to obtain. That reduces the threshold and cost for RISC-V developers and plays a huge role in promoting the ecosystem development of RISC-V. 

Looking beyond the obvious 

In a recent interview, Leiphone learned that some foreign netizens who evaluated the D1 were surprised that its performance surpassed the equivalent Arm chip. 

It was during the development of D1 that Sun Yanbang saw great possibilities for the CPU. 

“In the past, chips were made based on Arm’s cores. From the perspective of the system, there were actually many development needs that could not be met. However, when developing D1, there is more room to play. We not only increased the number of IP bits from 32 to 64, but also made a lot of optimizations in vector calculation and data throughput. Optimizing from a system perspective can bring a completely different experience to applications and ultimately to customers,” Sun Yanbang said. 

Fortunately, the D1 project provided the opportunity for Chen, who was working at T-Head to meet Sun at Allwinner Technology. Both saw that D1 could solve a huge dilemma for RISC-V which had no access to mass-produced AP chips. 

The two had the same vision for the future of RISC-V and believed that RISC-V could create the same ecosystem value as Linux. Chen, a Wenzhou native with entrepreneurial genes in his bones, recalled the teachings of his mentor Yan Xiaolang: “The chip is the most serious field in China right now. If the country has cultivated you, you should have the courage to devote yourself to this industry.” 

With an historic opportunity for a global open standard like RISC-V, Chen and Sun decided to start a large chip company with RISC-V architecture to fully promote the commercialization and development of RISC-V. 

Knowing that Chen and Sun wanted to start a business, Professor Gao Bingqiang was very supportive. Before the first version of the business plan was finalized he offered to provide angel funding support. 

Professor Gao told them: ”I haven’t invested in semiconductors for many years, instead I’ve been investing in applications. But the opportunity of RISC-V encourages me to invest in semiconductors once again. RISC-V is the future and there is an opportunity to create a world-class enterprise. This enterprise must also be in China because China has the most future-oriented applications. To make RISC-V succeed, we must have the custom design ability of the processor core and at the same time we must have the practical ability to land in the market. Your combination is the best.” 

Tang Lihua is the first executive of a listed chip company to pay any attention to RISC-V. He has always been optimistic about the commercialization and mass production of high-performance RISC-V CPUs. He too supports the entrepreneurial plans of his old friends Chen Zhijian and Sun Yanbang. 

Be smart, be strategic. 

RISC-V’s open standard gives all individuals and institutions an equal opportunity to develop CPUs. Many RISC-V players have focussed on developing and selling small-scale RISC-V IP cores, hoping to emulate Arm’s business model.  

However, the ecosystem development of RISC-V is still underpowered. What is required is a high-performance processor that can support large-scale software in this early development stage. Although everyone is optimistic about the development of RISC-V, most companies are still waiting to see what happens next. 

The financial resources of small companies can often only support research and development of small core IPs. Large companies are also conservative in their investment in RISC-V, unsure if it will bring huge commercial profits.  

The general shortcomings in the software capabilities of semiconductor companies makes things even more prohibitive. Nobody is willing to take risks when existing business revenue is stable. Therefore, even if there are players like T-Head who are willing to give up IP licensing fees in the short term and open source their RISC-V processors, there is still very little investment in high-performance RISC-V chips within the industry. 

Concerns about the commercialization of RISC-V often revolve around a simple problem: both X86 and Arm have established mature ecosystems and they dominate PC and mobile devices respectively. Does RISC-V really have a chance in the high-performance field? 

While public discussion is still stuck on whether RISC-V can break through on the PC, mobile or server side, SpacemiT has focused its attention on a more open future. 

According to Chen, ”Each generation of architecture develops symbiotically with new applications in the future. The X86 architecture has grown with the era of computers and the Internet, while the Arm architecture has become the mainstream architecture in mobile phones because of the mobile Internet. The next era will come from connecting people to things – the era of intelligence and interconnection of all things. From the Internet to the mobile Internet to the Internet of Things, the number of devices in each era continues to increase more than tenfold, At the same time, things become more open to the underlying technology requirements. The reality is that X86 and Windows in the Internet era are closed, Arm in the mobile Internet era is semi-closed and Android OS is semi-open. In the era of the Internet of Everything, both CPU architecture and OS hope to be completely open.” 

He added: “Different from the single device form of PCs and smartphones, the next era will present diversified scenarios and applications. The openness and DSA features of RISC-V are better placed to meet the requirements of the next era.” 

“Usually everyone’s understanding of chips in the era of the Internet of Everything may be IoT chips with low computing power. But this understanding may be a bit one-sided. The computing power requirements of intelligent IoT are varied. For example, a smart car is a typical scenario for intelligent IoT requiring very high computing power. In the next era, there will be more new application forms, such as intelligent robots, edge computing nodes, intelligent industrial and commercial scenarios, cloud computers and cloud mobile phones. With more open architecture and more diverse software, larger and more efficient computing power will be the trend for the next era.” 

“In the future, the computing power demand on the IoT side will inevitably increase, cloud computing power will inevitably sink, and the future computing power form and computing power structure will be reconstructed. That’s why SpacemiT’s RISC-V chips are aimed at new applications including intelligent robots, new edge computing nodes and the like. Future intelligent robots, humanoid robots and low-speed intelligent vehicles are all areas that we are optimistic about. Tesla’s release of the Optimus humanoid robot also reflects an understanding of the future world trend which coincides with our point of view. ” 

Chen also revealed the company’s current research and development progress. He says, ”The computing power of our first-generation products will exceed the performance of the original D1 by 200 times. The server-level RISC-V processor core project has also been launched while both chip and software development are in progress.” 

During his time at Allwinner Technology, Sun Yanbang participated in the cooperative development of several commercial products and witnessed the explosive growth and intelligence of sweeping robots. The primary form of the sweeping robot is a collision type, without the function of obstacle avoidance and path planning. When it encounters an item, it simply bounces back and continues to clean. This kind of product only needs an MCU-level main control chip. But current high-end sweeping robots have developed to the point where they require quad-core processors and 4T computing power requirements. In terms of function, they need to support multi-channel camera input, intelligent object recognition, path planning and obstacle avoidance. 

“The continuous growth in demand for computing power has become the eternal demand for electronic products,” said Sun Yanbang. “As we expand the scope of humanoid robots, the market targeted by SpacemiT, the product form has yet to fully emerge”.  

Sun is full of confidence in the future. He firmly believes that robots represent the intelligent future and will become one of the important markets for RISC-V. He says, ”We must plan ahead for the future – dominating the ecosystem is never something that can be grasped by latecomers.” 

Challenging application areas that have not yet been fully implemented potentially means higher uncertainty. The risks are as large as the opportunities. Chen and Sun’s bold entrepreneurial plans are regarded as ‘advanced’ by many people. But Chen remains firm on this: “I believe in what Steve Jobs said, if you think something is meaningful, do it immediately. If you work hard to make something truly meaningful to the industry and society, success is inevitable.” 

In order to respond quickly and prepare for these new opportunities, SpacemiT already has an IP R&D team, a chip system team and a full-stack software team. In order to speed up research and development, SpacemiT’s RISC-V IP core R&D and software development are carried out simultaneously. In a demo presentation to Leiphone, the SpacemiT team was able to successfully run large-scale computer software through the RISC-V chip platform, smoothly running websites and playing videos. 

According to Chen: “As the founder of a Chinese computing chip company, I am convinced that a world-class computing ecological enterprise will be born in China based on RISC-V architecture. Our vision is to grow into a computing ecosystem giant with the development of RISC-V. This is our time!” 

Reprinted from Leiphone