Zhongke Tongliang showcases its full-stack domestic RISC-V data-flow intelligent computing products at the 2026 CNCC China National Computer Conference
From October 21 to 24, CNCC 2026 China National Computer Congress and 2026 CCF Computer Exhibition was held in Chengdu. Zhongke Flux made a high-profile appearance with its SmarCo series products, showcasing the full-stack product matrix of domestic RISC-V data flow intelligent computing around the conference theme of "Data Convergence Drives Intelligent Computing for the Future".
Hosted by the China Computer Federation, CNCC has been successfully held for 22 sessions since its launch in 2003, making it a high-standard annual grand event in China's computing sector. This year's congress gathered 12 academicians and numerous national-level talents, with topics covering cutting-edge directions such as artificial intelligence, computing power infrastructure, big data, edge computing, and embodied intelligence. At such a high-standard academic and industrial event, the exhibition of Zhongke Flux has become a window to observe the technical route of domestic computing power and the trend of industrial implementation.
Zhongke Flux has long focused on data flow architecture, RISC-V heterogeneous computing and video intelligence. Its core team has more than 20 years of accumulation in video intelligence technology, won the global double championship of GreenGraph 500 large and small datasets, obtained more than 140 cumulative patent authorizations, and R&D personnel account for more than 50% of its total staff. The company is committed to building a fully independent and controllable intelligent computing base for software and hardware, providing efficient and low-latency computing power support for high-concurrency data scenarios.
During the exhibition, the booth of Zhongke Flux always maintained high popularity. Especially in front of the real-time demonstration area of AI workstations and AI Digital Employee Box, batches of visitors stopped to watch and experience hands-on. The technical staff frequently communicated with visitors on issues such as RISC-V ecology, multi-channel video processing, and edge deployment, and many users from universities, scientific research institutes and enterprises expressed their intentions for subsequent testing and cooperation on site.
At the chip and module level, Zhongke Flux demonstrated the SmarCo GC3 chip and the SmarCo Rocket AGX GC series intelligent computing modules. The GC3 chip adopts a native data flow architecture, positioned as "born for video intelligence", and provides a domestic computing power solution with both high concurrency and low latency in scenarios such as video processing, large model inference and edge intelligence. The Rocket AGX GC series intelligent computing modules further encapsulate the chip capabilities into a deployable form, which can easily handle heavy workload tasks such as large model inference, multi-channel high-definition video analysis, complex machine vision and real-time decision-making, featuring high energy efficiency ratio, high computing power utilization and low latency. In front of the booth, many visitors specifically consulted on the data flow architecture and power consumption performance of the GC3 chip, as well as the interface compatibility, heat dissipation design and mass deployment scheme of the modules, and the technical staff answered them one by one combined with application cases.
In terms of desktop products, the SmarCo AI workstation exhibited by Zhongke Flux became one of the focuses on site. This product is positioned as a desktop high-performance intelligent computing terminal, specially built for scenarios such as video intelligence, large AI model, and robot R&D and debugging. Adopting data flow architecture, it features high concurrency and low latency, provides 250 TOPS INT8 super large computing power, 128GB LPDDR5 ECC unified memory, real-time hardware decoding capability for 128 channels of 1080P high-definition video, and integrates 12 RISC-V cores. This workstation emphasizes "full interface opening, zero threshold for scientific research debugging", and provides rich expansion interfaces to meet the secondary development and debugging needs of scientific research institutions, universities and enterprise developers.
At the application level, the SmarCo AI workstation covers scenarios such as local large model deployment and development, AIGC video generation, real-time video AI analysis, edge computing, embodied intelligence, and AI4S. During the congress, Zhongke Flux conducted live real-time demonstrations at the booth, allowing visitors to intuitively feel the operation performance of desktop-level intelligent computing terminals in real tasks. During the demonstration, the technical staff ran local large model inference and video analysis tasks in real time. A participant from the computer college of a university came to the booth for two consecutive days, focusing on inquiring about the openness of the scientific research debugging interface; a participant engaged in embodied intelligence research said that the desktop-level device can provide 250 TOPS computing power and 128GB unified memory, which is very attractive for laboratory algorithm verification; another technical person in charge of a security enterprise focused on inquiring about the real-time hardware decoding capability of 128 channels of 1080P video, and agreed to conduct further tests after the meeting.
Another product that carried out live real-time demonstration is the SmarCo AI Digital Employee Box. This product is positioned as a lightweight edge AI terminal, with the core goal of "turning every existing screen into an intelligent service entrance". Its features include upgrading existing screens, local secure intelligence, active perception and instant service, enterprise knowledge precipitation, and flexible adaptation to diverse spaces. Application scenarios cover automobile 4S stores, retail exhibition points, service counters, communication business halls, bank outlets, etc. Compared with cloud-based solutions, edge AI terminals have obvious advantages in data localization, response real-time performance and deployment flexibility, and are especially suitable for offline scenarios with high requirements for security and instant services.
In the live demonstration, the AI Digital Employee Box demonstrated its active perception and instant service capabilities. A visitor stood in front of the screen to simulate customer consultation, and the Box responded quickly, triggering a heated discussion around. A person in charge from a communication business hall asked in detail about the knowledge base update method and multi-store unified management scheme, and added the WeChat of the staff on site, hoping to arrange a pilot later. In addition, people from retail, banking and other industries had in-depth exchanges with the staff on issues such as deployment cost, existing screen adaptation and local data security. Many visitors took out their mobile phones to shoot the demonstration process, and the on-site interaction atmosphere was very warm.
From the perspective of product matrix, Zhongke Flux has built an intelligent computing product system from chips, accelerator cards, modules to workstations and edge terminals, forming synergistic effects in dimensions such as localization, low latency and high energy efficiency. Under the conference theme of "Data Convergence Drives Intelligent Computing for the Future", this full-stack layout has become a highlight at the CNCC exhibition this year.
At present, with the accelerated implementation of applications such as large models, video intelligence and embodied intelligence, the demand for computing power is extending from the centralized cloud side to the edge side and desktop side. How to realize efficient, secure and controllable computing power supply in different scenarios has become a common problem faced by the industry. Zhongke Flux said that it hopes to conduct in-depth exchanges with partners from academia and industry through the CNCC platform, and jointly promote domestic intelligent computing technology to a wider range of application scenarios.