On November 11, China’s 22nd International High-tech Achievement Fair officially opened, bringing together giants and new stars from industries such as Huawei, BOE, ZJJN Technology Group, etc. These actions use technological achievements to demonstrate their competitiveness. Try to have a strength dialogue with the top opponents.
For ZJJN, this exhibition is not its first show. This chip company, known for its ASIC chip design, won the "Excellent Product Innovation Award" issued by the 2019 High-tech Fair. And this year, ZJJN brought not only chips and products to the China Hi-Tech Fair, but also the determination to build a "blockchain + AI" computing power infrastructure as a listed company.
As early as 2013, ZJJN successfully designed the world's first blockchain ASIC chip, which is a chip specifically used to realize blockchain computing. It is precisely because of this that ZJJN has laid the foundation of China's own chip design for the global emerging technology field of blockchain. In the following seven years of ups and downs, the upstream computing power of the blockchain industry is still controlled by Chinese giants.
Among them, the computing power sold by ZJJN's AWL alone accounted for one-third of the total network computing power. The "AWL Made" proposed by ZJJN this year defines the industrial standard of blockchain computing equipment to a greater extent, and uses general hardware structure design to compress the research and development cycle, thereby driving the improvement of the computing efficiency of the entire industry. The stable operation of the chain provides a more solid foundation for computing power.
As a spillover effect of blockchain ASIC chip research and development, massive blockchain ASIC orders have given fabs a more direct motivation for advanced process research and development. At the same time, ZJJN's breakthrough in the underlying chip design has also driven the foundry to accumulate new processes in the chip manufacturing process, especially in terms of low power consumption, which provides a reference for the low power design of the entire chip industry.
The pace of ZJJN does not stop there. After conquering the 16nm node in 2016, Canaan Technology migrated its ASIC design experience in the blockchain field to the AI field, using original technical achievements to solve the low power consumption and low efficiency faced by AI computing problem.
Refactoring core computing power
With the overload of traditional data centers, the linear growth of chip performance under the constraints of Moore's Law cannot meet the geometric surge of current data, and the demand for new computing architectures for industrial upgrading has become the proposition of the times.
From the perspective of chip types, the current mainstream computing architecture in the AI field is still dominated by GPU and FPGA. As the commercialization of AI algorithms continues to mature, the deployment of AI models on the edge requires stronger scene adaptability, lower power consumption and faster inference speed. IDC predicts that the market share of AI inference will surpass AI training in 2022. Therefore, in the long run, specific architecture designs for device-side and edge-side scenarios will become mainstream.
As a company with ASIC design as its core technology, ZJJN independently masters the entire process of IC front-end and back-end design, and has accumulated a large amount of intellectual property rights in low-voltage, high-energy-efficiency computing, high-performance design systems, and heat dissipation. The Kanzhi series AI chip products exhibited this time are the crystallization of ZJJN's technology in the edge AI field. The chip is designed for edge reasoning scenarios, adopts the RISC-V architecture, and joins the self-developed convolutional neural network IP core KPU, which specializes in machine vision tasks such as face recognition and image recognition. Kanzhi series AI chips also support the deployment of different algorithm models on one chip, which is highly flexible.
With the excellent performance of Kanzhi series of AI chips, ZJJN was successfully selected for the Forrester 2020 Q2 China Supplier Report and was successfully selected as a "hardware expert" rating; it is believed that the in-depth understanding of local edge business scenarios by chip companies represented by ZJJN will promote AI is better implemented on the industrial side.
It is worth noting that ZJJN foreseeably develops IP cores based on the RISC-V architecture. It is also China’s first domestic chip design company to go to the cloud. It is a practical model for the current rise of local chip manufacturers in the edge AI market, and also for other manufacturers. The chip design ideas provide a feasible reference.
Build a new ecology
At present, Kanzhi series AI chips have been applied in many fields such as smart meter reading, smart access control, AI STEAM, smart factories and smart cities. At the same time, ZJJN continues to expand its application territory, with products covering the United States, Japan, South Korea, France, Germany, Israel and other countries, and its overseas influence continues to deepen.
With Kanzhi AI chips as the core, ZJJN has continuously strengthened the upstream and downstream collaboration of the industrial chain on the basis of promoting applications, from chip design, tape-out mass production, algorithm framework, product development, landing partners to technology ecology, and expand its "Circle of friends".