The first | zhongke wuqi pre A+ round of financing of hundreds of millions of yuan, five years of self-built Internet of Things version of Wintel

As early as 2012, the Chinese Academy of Sciences launched the information technology pilot project, laid out the research and development of the next-generation information infrastructure IT3.0, and opened a strategy called "sea cloud computing".

Among them, the team led by Chen Yunji and Dr. Chen Tianshi deeply cultivated "cloud computing" and founded Cambrian, which developed server-oriented acceleration chips, and later became the first ai chip on the science and technology innovation board, which attracted much attention in the industry.

In the field of "sea computing", that is, the field of object-end computing, it is a team led by Dr. Zhang Lei and Dr. Wang Yuantao to achieve local acceleration and interconnection for massive terminal equipment, until 2018, with the strong support of the Institute of Computing of the Chinese Academy of Sciences, in the name of habitat technology, the Zhongke Cohabitation Company for the Internet of Things was founded. Although the Chinese habitat is not as well-known as the Cambrian, it is also promising.

On May 10, Zhongke Wuqi announced the completion of nearly 300 million yuan Of Pre-A+ round financing, jointly invested by Nanjing Qilin, Zhongke Advanced, Zhongke Turing, National Science and Technology Achievement Transformation Guidance Fund, Saifu and so on. Zhang Lei, founder of Zhongke Wuqi, said that this round of financing will be mainly used for the research and development of core technologies, innovative application scenarios and business development.

The earliest start-up team in China to contact RISC-V, and Cambrian "out of the same door"

Starting from the Chinese Academy of Sciences, Zhongke Wuqi did not choose the ecologically perfect X86 and Arm architecture, but chose the immature RISC-V, which is one of the earliest start-up teams in China to contact RISC-V.

The choice of RISC-V instruction set architecture is closely related to the work experience of founder Zhang Lei in the Institute of Computing, Chinese Academy of Sciences.

"For many years, we've been working on system-on-chip and hopefully innovating, but whether it's x86 or Arm, the intellectual property is owned by the company, and if you want to change it, you have to pay for it, which is not friendly to academic research, and it is difficult to achieve innovation without sufficient funding." For this reason, my research group has been looking for an open source instruction set. Zhang Lei introduced.

Open RISC is the first open source instruction set that Zhang Lei et al. tried, this instruction set is provided by the open source organization OpenCores, but then its parent company ORSoC AB was in a profitable position in 2015 to consider shifting the focus of business to miners, the Open RISC community ecology is declining due to unmanned maintenance, at the same time, David Patterson of the University of California, Berkeley is pushing away the new instruction set architecture of source openness. After struggling for a while, Zhang Lei, who was keen on Berkeley's influence, finally switched from Open RISC to RISC-V with his research team.

After that, when the fourth RISC-V symposium was held at MIT, Zhang Lei also made a relevant report at the conference, and was the only Chinese in the audience. After returning to China, in the case that the industry generally did not understand what RISC-V was, Zhang Lei promoted the Chinese Academy of Sciences to join the RISC-V International Foundation, and the Chinese Academy of Sciences has since become the earliest organization in China to join the RISC-V Foundation.

By 2017, members of Zhang Lei's research team began to try to develop chips with RISC-V for the first time. "We were bullish on RISC-V, so we thought about making a quick one to try." Zhang Lei recalled. So Zhang Lei and others used the open source code in the community to add their own designs and design a CPU based on RISC-V.

This chip uses the 55nm process, which is one of the earliest chips in China to adopt the RISC-V instruction system, has a heterogeneous dual-core processor, and integrates the programmable AI dedicated accelerator NPU.

In May 2018, after the test of the first RISC-V CPU was completed, Zhang Lei and his team officially founded Zhongke Wuqi and began to develop the second chip.

In May 2019, the second generation of JX2 tape-out film of Zhongke Wuqi was successful, which is the first RISC-V chip that can run Linux systems in China, supporting more than 20 mainstream AI algorithm models.

After the success of the JX2 tape-out, Zhang Lei and others found that the ecologically unsound RISC-V had problems such as software adaptation, so instead of mass-producing JX2, they continued to develop JX3 and adopted the RISC-V + Arm hybrid architecture to achieve mass production in 2021.

Five years of precipitation, software and hardware decoupling to cope with fragmentation

What makes people curious is that as two companies in the same special project of the Institute of Computing of the Chinese Academy of Sciences, why is Zhongke Wuqi still in the growth stage?

Zhang Lei explained to Leifeng Network that Zhongke Wuqi is facing a large number of fragmented smart devices, and the use scenarios of Cambrian are more clear and concentrated, so Zhongke Wuqi has no way to concentrate on outbreaks like Cambrian and quickly open the market.

"Wuqi is facing a large number of end-of-the-thing equipment and needs to deal with the challenge of fragmentation, so we must standardize our platform, technology and product quality before we can bring it to the market, and at the same time, we need customer feedback, which takes a certain amount of time." Zhang Lei said.

For the fragmentation of the Internet of Things, Zhang Lei believes that the fragmentation of the Internet of Things is only a phenomenon, not a problem, can not change it, can only adapt to it, so in the past five years, Zhongke Wuqi has been looking for solutions to deal with fragmentation.

In fact, the entire computer system is divided into four parts, processor, memory, communication, input and output modules. Because different IoT devices have different ways to obtain information, or use control motors, or use cameras to collect image information, or use other types of sensors, so the input and output modules have become the largest source of IoT fragmentation, and the other three modules have a certain commonality as parts that every IoT device must have.

"So we standardize the common part, put the chip memory, Flash, Wi-Fi, etc. on a thumb-sized machine, equipped with different sizes of computing power, memory, and make three versions of low-end, medium-matched, and high-end for customers to choose and leave personalized things to customers."

"Thumb-sized machine" is the self-developed material-end ultra-microcomputer "Hermit Crab", "Hermit Crab" integrates the "JX" series chip developed by Zhongke Wuqi, pre-installed Wuqi OS intelligent operating system, and adopts 2.5D/3D packaging technology to integrate the storage and communication systems at the bare die level, thereby condensing a complete computer core system into an ultra-small size All-in-one machine that is only the size of an SD card.

In addition, Wuqi's ultra-microcomputer and peripheral devices adopt a standard pluggable interface mode, which can connect various sensors and actuators through such as PWM, I2C, UART, GPIO, USB, SDIO, SPI, MIPI, etc., which means that plug-and-play can be realized on The Internet of Things device can be implemented and reduced the threshold for development and innovation of hardware products.

In addition to the standardization of the hardware level, Wuqi also starts from the software level to achieve software and hardware decoupling, that is, to let the software and hardware develop according to their respective iteration speeds.

Zhang Lei said that people who do hardware now will also define the software on the hardware, such as Xiaomi's smart socket, the software and hardware are all made by Xiaomi itself, but the software and hardware on the phone are completely different, and the software is completely supported by a third party. "According to the logic of mobile phones, after the decoupling of software and hardware, the problem of fragmentation of the Internet of Things can be solved to a large extent."

Because of this, Zhongke Wuqi launched the Wuqi OS operating system, decoupling hardware devices and software applications, making IoT devices an open platform for third-party applications, thereby realizing the expansion of functions.

Leifeng network learned that the Wuqi OS includes three core parts: device open platform (hardware access), third-party developer platform (small application development) and Wuqi bubbling (user entrance). The device open platform provides bare metal, Linux and AOSP kernel versions for end-of-thing devices with different capabilities, while the third-party developer platform is oriented to the application market of end-of-world devices, supporting third-party developers to implement "human-machine" intelligent application scenarios in low code, and the bubble is the entrance for end users to enter the material-based OS, connect control devices and obtain information.

Copy the computer, mobile phone paradigm, self-built Wintel in the era of the Internet of Things

"The original intention of Theosophy is not a company that only makes chips, what we want to do is chip plus operating system, but the order of research and development is the first chip, the operating system is behind, and the chip research and development is more troublesome, slowly exposing the rhythm problem." Zhang Lei said.

Wuqi firmly believes in the path he has chosen, believing that the value chain of information systems is transmitted from top to bottom from beginning to end.

Intel is almost one of the most valuable companies in the PC era, in the final analysis, in addition to Intel's own strength, the most important thing is the Wintel alliance built by Microsoft and it, the Windows operating system has a large number of user-oriented applications including office, browser IE, etc., so Intel's value is further increased.

Similarly, the top-down value chain is also staged in the era of mobile intelligence, giving birth to the arm + Android strong binding combination.

Because of this, Zhang Lei realized that no matter how well the RISC-V AI chip is, without a strongly bound operating system and derived application software, it will not have absolute value in the end.

"So we first made a chip with a relatively long cycle, and in fact, the overall plan was from the chip to the operating system to the end user." Zhang Lei said.

At present, Zhongke Wuqi has initially realized the original planning, built a set of solutions from chips to operating systems to ultra-microcomputers, and has been applied in various scenarios such as home, medical, sports, health, and integrated media.