Author | Yunpeng Editor | Moying
As AI accelerates its shift from the cloud to the edge and endpoint, Agentic AI is rapidly penetrating various industries. A series of physical industries, including transportation, oil and gas, robotics, healthcare, and manufacturing, have been swept into the wave of intelligent transformation. Computing power support at the edge and endpoint has become a critical cornerstone driving the digital and intelligent transformation of the physical world. Such changes have brought a series of new challenges: industrial control scenarios require balancing microsecond-level real-time control with multi-task concurrency; edge gateways need to handle multi-protocol conversion and data preprocessing simultaneously; and machine vision requires completing rapid AI inference on-site.
Overall, the demands for real-time performance, reliability, data security, and bandwidth costs are driving the transformation and sinking of computing power. New embedded computing needs are constantly emerging, and the demand for domestic platforms in massive embedded scenarios is growing higher.
Under this trend, Hygon Information Technology Co., Ltd. (Hygon) officially released the Hygon 1000 series CPU—Hygon C86 architecture processors built for low-power, embedded, and endpoint computing power needs. Featuring high energy efficiency, long lifecycle, native compatibility, and intrinsic security, it can provide stable and reliable basic computing power for edge terminals, industrial equipment, and IoT (Internet of Things) nodes in the integrated "cloud-edge-endpoint" computing power system. Meanwhile, Hygon's embedded ecosystem construction and software ecosystem system are developing rapidly. A series of ecosystem-level support measures were released simultaneously, and we have seen its partners actively practice and implement a series of excellent cases. This is a step for Hygon to expand its chip territory and enter the new embedded market, and also a critical step for domestic chips to expand their computing territory from data centers to physical industries.
01.Mature Technology and Ecosystem of CPU and DCU Products Solid Foundation in High-End Computing in Data Centers
Hygon is no "novice" in entering the embedded field. Its computing products such as CPUs and DCUs have long laid a solid foundation in the high-end computing field of data centers, and these accumulations have become key support for this territorial expansion. Its CPUs are built on the C86 instruction set architecture, compatible with the x86 ecosystem, focusing on smooth migration and low-cost deployment. From the data center-level Hygon 7000 series to the mid-to-high-end commercial Hygon 5000 series, it caters to both high-computing-intensive scenarios and scenarios that pay more attention to energy efficiency and cost-effectiveness.
The DCU (Deep Computing Unit) belongs to GPGPU, specifically designed for parallel computing and AI (Artificial Intelligence) acceleration. From Zhisuan No. 1 filling the domestic gap, to Zhisuan No. 2 comprehensively adapting to the training and inference of mainstream large models such as ERNIE Bot, Tongyi Qianwen, and LLaMA, the DCU software ecosystem is now tending towards perfection. It is compatible with the "CUDA-like" development environment and natively supports mainstream domestic and foreign deep learning frameworks and computing libraries. In terms of commercial implementation, Hygon's CPU and DCU products have become one of the main domestic computing power foundations in the cloud data of several major telecom operator giants, and have been implemented in multiple domestic research institutes, advanced computing centers, and intelligent computing centers in various regions.
Over the years, the technological accumulation and C86 ecosystem moat established by Hygon in the fields of data centers and high-performance computing are the key support for its ability to leverage advanced capabilities to expand into embedded scenarios. On the one hand, the extreme stability and security required by data centers are crucial for industrial and embedded scenarios. On the other hand, years of deep cultivation have given Hygon a deeper understanding of AI application scenarios and a more precise grasp of the real pain points of AI implementation. The accumulation of extensive know-how makes Hygon more at ease when entering the embedded AI track for the first time.
02.Hygon 1000 Targets High Energy Efficiency and Lightweight Design Advancing Hand in Hand with Software and Hardware Ecosystems
Based on its profound accumulation in high-end computing scenarios in data centers and the mature product and technology ecosystem of CPUs and DCUs, Hygon has further expanded its chip capabilities to low-power, embedded, and endpoint scenarios, extending its application scope to robotics, smart devices, and edge computing and other areas closer to the physical world.
The newly released Hygon 1000 series targets the embedded market. It adopts the C86 architecture, is positioned as a lightweight CPU, and is the first product with integrated display functions, compatible with mainstream software and hardware ecosystems in the market. The Hygon 1000 can serve as a computing platform for slim, fanless cloud desktops, supporting dual-screen display and adapting to applications such as cloud terminals and educational terminals; serving as the main control for industrial robots and AGVs, synchronizing sensors and motors through TSN real-time networks; embedded in production line touch screens and vision controllers, to solve on-site computing needs such as industrial control and visual inspection; and can also be deployed in edge and outdoor devices to process sensor data nearby, reducing the network pressure caused by continuously uploading data to the cloud.
Overall, the Hygon 1000 focuses on high energy efficiency and lightweight design, covering a very wide range of scenarios. In addition, Hygon revealed that in the future, it will carry out new product R&D for different computing needs such as mainstream commercial, endpoint AI, and high-end professional applications. Behind the realization of such extensive scenario coverage is the solid technological innovation support of the Hygon 1000 series CPU.
Its C86 architecture is compatible with mainstream software and hardware ecosystems, supporting API interfaces such as DirectX, OpenGL 3.3, and OpenCL 3.0, with good compatibility and adaptability in various scenarios. Existing industrial software can be smoothly migrated without major code reconstruction, greatly reducing the risk cost of users' secondary development and R&D.
In the embedded industry, a large amount of industrial control, industry applications, and professional software have long lifecycles. Therefore, software and hardware ecosystem compatibility is an important foundation for the long-term deployment of embedded computing platforms. At the hardware architecture level, the Hygon 1000 series features a 4-core, 8-thread design, supports 4K 60FPS dual-screen display and DDR4/DDR5 memory, has a typical TDP as low as 10W, and possesses -40°C to 105°C wide-temperature capabilities. In addition to the two outstanding advantages of low power consumption and wide temperature, the Hygon 1000 series' industrial-grade long-term stable operation, long lifecycle guarantee and continuous supply capability also directly address the pain points of embedded scenarios such as edge terminals, industrial equipment, and IoT nodes. Support for fanless and passive cooling designs can reduce the failure rate, and support for ECC and hardware cryptographic acceleration security computing units can enhance security. Hygon has a 10-year product lifecycle commitment for the 1000 series CPU, while also setting up long-term supply plans, advance EOL notification mechanisms, and multi-generation compatibility and smooth upgrade mechanisms.
In industries such as industrial, energy, and transportation, equipment usually has a long lifecycle, with high requirements for continuous operation and complex deployment environments. At the same time, chips in the embedded field often need not only to meet current performance needs but also to match the long development, deployment, and maintenance cycles of industry equipment. The Hygon 1000 series can well meet these needs and solve customers' worries. Of course, for any chip product to truly achieve commercial success, the ecosystem is crucial, especially in the embedded market. Once customers choose a platform, the switching cost is extremely high, making the software and hardware ecosystem the fundamental guarantee. Hygon's value positioning in the embedded market is very clear: it does not just sell single chips, but hopes to become a "long-term partner for ecosystem, security, and stability". Focusing on the embedded industry, Hygon coordinates the ecosystem among chips, complete machine ODM/OEM partners, operating systems, software partners, ISVs, and industry users, providing support such as adaptation certification, joint solutions, talent cultivation, and market docking.
Overall, the Hygon 1000 series has a precise grasp of the pain points of implementing AI in the embedded field. Its core upgrades and feature focuses are used on the "cutting edge". It not only has a solid hardware foundation but also advances hand in hand with the software ecosystem, solving the core issues customers care about and becoming the "better solution" for CPUs in the embedded field.
03.From Data Centers to Edge and Endpoints Domestic Chips Accelerate Towards Full-Scenario Coverage
Expanding from data centers to the edge and endpoints, we see that the changes in Hygon's chip layout are highly consistent with the development trend of the entire AI computing industry. This move has deeper significance and impact on Hygon's own chip business and industry development. At the technical level, this is obviously a test of technology. Whether the products and technology ecosystem built on the C86 architecture can cope with embedded market demands and well solve pain points is a verification of implementation feasibility. At the business level, this expansion enables Hygon's chip business to better cope with the more fragmented, complex, and highly challenging edge and endpoint AI demands in the current AI era. In the past, the data center market had high customer concentration, large single-order scale, and relatively standardized demands. In contrast, in the embedded/endpoint market, demands are highly fragmented, scenarios vary greatly, customization levels are high, and the number of customers is extremely large.
To tackle this "hard bone", not only do the products and technology need to be excellent, but Hygon also needs to establish a rapid response system capable of supporting massive embedded scenario customers. At the customer level, expanding from relatively concentrated cloud computing and large government and enterprise customers to highly dispersed industrial OEM/ODM customers, system integrators, and ISVs, the expansion of the customer base and the increase in ecosystem coverage breadth are evident. For Hygon, "expanding from the center to the endpoint" is not only a production line expansion but also an expansion of Hygon's business boundaries, an upgrade of customer structure, and accelerated growth of the technology ecosystem. This means the implementation and perfection of Hygon's "cloud-edge-endpoint" full-scenario computing power ecosystem. They are not just a "chip supplier" but an "ecosystem co-builder".
In the embedded field with extremely high ecosystem barriers, Hygon has truly broken through the "last mile" from chips to terminal products through full-stack software and hardware collaboration and partner support plans. For the domestic chip industry, the release of the Hygon 1000 series means that domestic chips can cover broader computing scenarios, further enhancing their competitive ability in full-scenario and full-dimensional competition with international giants in the global market. It also gives enterprises more choices and more cost-effective and secure domestic computing power solutions.
04.Conclusion: Consolidating Technological Foundations and Expanding Ecosystems Horizontally The Digital and Intelligent Transformation of the Physical World is a Marathon
The sinking of computing power is an important part of the digital and intelligent transformation of the physical world. The release of the Hygon CPU 1000 series further promotes the development of this trend, fills in Hygon's territory in the low-power and embedded fields, and injects new vitality into the C86 architecture at the edge and endpoints. "Technical accumulation at the center feeds back to the endpoint, and the prosperity of endpoint scenarios feeds back to the center ecosystem," the formation of this virtuous cycle will also give Hygon more sustainable commercial vitality.
As the world's largest manufacturing and industrial automation market, China has the richest embedded application scenarios globally. The intelligent transformation and development of enterprises in these scenarios require solid, trustworthy, and long-term endpoint computing power foundation support. The digital and intelligent transformation of the physical world is a long-distance run, a comprehensive test of products, technology, ecosystem, and services. The expansion of domestic chips in the embedded field has consolidated the computing power foundation for the security, autonomy, and sustainable development of the digital and intelligent transformation of China's real economy.