Domestic computing power is on the rise! China's "invisible trump card" in the era of intelligence is starting to make its mark
Today, as the digital economy and the era of intelligence converge, computing power has become the core infrastructure for national competitiveness and corporate innovation.
However, against the backdrop of increasing uncertainty in the globalized supply chain, "autonomy and controllability" is no longer an option, but a necessity.
Longjing Technology, in collaboration with domestic technology enterprises such as Huawei and Feiteng, has made solid product capabilities and systematic layouts. With the vision of "building an intelligent world of interconnected everything", it has quietly established a "fully domestically produced base" for China's computing infrastructure.
From GPU servers to silent liquid-cooled workstations, from Feiteng, Haiguang to Kunpeng processors, from data center-level servers to portable workstations, Longjing has achieved vertical integration of "chip-hardware-scenario" with a series of high-performance, high-reliability, fully self-developed products.
01
product matrix
Longjing Technology focuses on three dimensions: computing power density, energy efficiency ratio, and scenario adaptability,
Build a complete solution from the data center to the edge.
High-performance GPU server: Targeted at AI training and high-performance computing
LG 4215G-10FA is a 4U dual-socket 10-GPU server launched by Longjing, based on the Feiteng S5000C-64-core processor. It supports 10 GPU cards, is fully compatible with PCIe 5.0 protocol, and boasts a 75% increase in memory bandwidth and a 100% increase in transmission rate compared to its predecessor. It is suitable for AI training, high-performance computing, and big data analysis scenarios.
The LH 4402GN-8KA, with the same positioning, is based on four Kunpeng 920 processors and comes standard with eight Ascend 910 chips. Its FP16 computing power reaches 2.5 PFlops, equivalent to the AI processing power of 500 traditional servers. It is suitable for high-precision computing scenarios such as FP32 and FP16, especially for large-scale model training and inference.
This model has been applied to the recommendation algorithm training of a leading Internet plus enterprise, compressing the model iteration cycle from 72 hours to 18 hours.
Tower server: An enterprise-level choice for flexible deployment
LG 5213T-HA is a tower server based on the Haiguang 7000 series processor, supporting dual CPUs, six PCIe 4.0 expansion cards, and 16 DDR4 memory slots. It boasts excellent scalability and compatibility. The tower design makes it an ideal choice for virtualization deployment and high-performance office computing in small and medium-sized enterprises. It is a typical representative of the "integrated hardware and software" localization alternative and has been widely used in fields such as financial databases and government cloud platforms.
Silent liquid-cooled workstation: An "invisible weapon" for edge inference
LD 5112W-2HA is an innovative product from Longjing in the field of edge computing and AI inference. It supports 560 TOPS INT8 computing power provided by two Atlas 300i NPU cards, features portability and silent design, and utilizes a customized liquid cooling system to achieve "extremely low noise" operation. This addresses the pain points of traditional AI devices, such as high noise levels and difficult deployment, making it suitable for noise-sensitive scenarios such as laboratories and offices. It has become the preferred device for AI localization applications in small and medium-sized enterprises (SMEs) ranging from 70 to 32b.
02
hardcore strength
Longjing Technology has established a clear "fully domestic" technology path, covering the entire stack of processors and achieving autonomy in GPU/NPU. It has achieved comprehensive computing power support from ARM to x86 architecture. The adoption of customized fans and liquid cooling systems in multiple products is not simply a matter of component assembly, but rather a deep technological integration and reinvention.
Fully-stack, autonomous and controllable chip architecture
Longjing has deep cooperation with chip manufacturers such as Feiteng and Kunpeng, optimizing BIOS and drivers for AI workloads.
Taking LH4402GN-8KA as an example, the processor affinity scheduling algorithm has increased the utilization rate of the computing power of the Ascend 910 NPU from 70% to 92%, resulting in a 31% increase in processing speed in image recognition tasks.
At the CPU level, we cover three major domestic chip systems: Feiteng (ARM architecture), Haiguang (x86 architecture), and Kunpeng (ARM architecture). In terms of acceleration chips, we adopt dedicated AI chips such as Ascend 910/NPU and Atlas 300i to achieve autonomous and controllable computing power supply.
Global application of liquid cooling technology
From data center models to portable workstations, Longjing has miniaturized and customized liquid cooling technology. Through structural optimization, customized fans, and liquid cooling technology, it has improved heat dissipation efficiency while achieving a 15% reduction in power consumption and an 80% reduction in noise, truly realizing "cool operation and green computing".
Server level: Increase the air intake area by 15% and enhance the heat dissipation efficiency by 50%;
At the workstation level: the volume is reduced by 40%, and the noise level is controlled below 45 decibels (equivalent to a library environment).
Hardware and software co-optimization
Longjing has implemented designs such as "tool-free maintenance", "out-of-band visual management", and "organic log automatic recording" across its entire product line, significantly reducing operational complexity. From hardware monitoring to fault warnings, from log analysis to remote maintenance, Longjing has established an intelligent service system that covers the entire product lifecycle, allowing customers to "buy with confidence and use with peace of mind".
The self-developed "Longzhi" management platform can monitor the health status of hardware in real-time.
Longjing's products are genuinely supporting the intelligent transformation of various industries in China.
03
Empowerment by large models
For enterprises, localization is merely the foundation; the key lies in the ability to rapidly generate value. The Linglong Digital Intelligence Cloud Platform (AIOC), jointly developed by Shanghai Longjing Technology and Shanghai Supercomputing Center, boasts three major modules: "Wenshu - Yanhai - Zhixi". Targeted at enterprise-level users, it truly addresses the pain point of the "last mile" in AI implementation.
Wenshu: An intelligent assistant deeply integrated with a knowledge base, supporting regulation inquiry and fault diagnosis, enabling efficient access to key information.
Yanhai: A powerful auxiliary tool that integrates intelligent writing, polishing, translation, and question-answering, significantly enhancing content creation efficiency.
ZhiXi: Under the premise of ensuring data security and privacy, it provides accurate document interpretation, intelligent question answering, and multidimensional knowledge guides.
The AIOC platform, based on a domestic AI computing framework, can operate stably on various domestic CPU servers and operating systems after deep optimization. It supports remote installation and local deployment, helping enterprises build a privatized AI service system that ensures "data remains within the domain and computing power is independently controlled", meeting the stringent requirements for data privacy and security compliance in key areas such as government affairs, finance, energy, and military industry.
Finally,
When the LD5112W-2HA workstation enters an ordinary office, and when the LG4215G-10FA supports the training of national-level AI large models,
These "Chinese cores" are redefining the boundaries of computing power.
Longjing Technology provides a secure, efficient, and green computing engine for various industries with its integrated solution of "hardware + software + services".
When we open the chassis of Longjing servers, what we see is not only a dense array of chips and circuits, but also the determination of Chinese technology enterprises to break through blockades. From the optimization of the instruction set of Feiteng processors to the fluid mechanics design of liquid cooling systems, these "invisible innovations" are reshaping the global computing industry landscape.
However, against the backdrop of increasing uncertainty in the globalized supply chain, "autonomy and controllability" is no longer an option, but a necessity.
Longjing Technology, in collaboration with domestic technology enterprises such as Huawei and Feiteng, has made solid product capabilities and systematic layouts. With the vision of "building an intelligent world of interconnected everything", it has quietly established a "fully domestically produced base" for China's computing infrastructure.
From GPU servers to silent liquid-cooled workstations, from Feiteng, Haiguang to Kunpeng processors, from data center-level servers to portable workstations, Longjing has achieved vertical integration of "chip-hardware-scenario" with a series of high-performance, high-reliability, fully self-developed products.
01
product matrix
Longjing Technology focuses on three dimensions: computing power density, energy efficiency ratio, and scenario adaptability,
Build a complete solution from the data center to the edge.
High-performance GPU server: Targeted at AI training and high-performance computing
LG 4215G-10FA is a 4U dual-socket 10-GPU server launched by Longjing, based on the Feiteng S5000C-64-core processor. It supports 10 GPU cards, is fully compatible with PCIe 5.0 protocol, and boasts a 75% increase in memory bandwidth and a 100% increase in transmission rate compared to its predecessor. It is suitable for AI training, high-performance computing, and big data analysis scenarios.
The LH 4402GN-8KA, with the same positioning, is based on four Kunpeng 920 processors and comes standard with eight Ascend 910 chips. Its FP16 computing power reaches 2.5 PFlops, equivalent to the AI processing power of 500 traditional servers. It is suitable for high-precision computing scenarios such as FP32 and FP16, especially for large-scale model training and inference.
This model has been applied to the recommendation algorithm training of a leading Internet plus enterprise, compressing the model iteration cycle from 72 hours to 18 hours.
Tower server: An enterprise-level choice for flexible deployment
LG 5213T-HA is a tower server based on the Haiguang 7000 series processor, supporting dual CPUs, six PCIe 4.0 expansion cards, and 16 DDR4 memory slots. It boasts excellent scalability and compatibility. The tower design makes it an ideal choice for virtualization deployment and high-performance office computing in small and medium-sized enterprises. It is a typical representative of the "integrated hardware and software" localization alternative and has been widely used in fields such as financial databases and government cloud platforms.
Silent liquid-cooled workstation: An "invisible weapon" for edge inference
LD 5112W-2HA is an innovative product from Longjing in the field of edge computing and AI inference. It supports 560 TOPS INT8 computing power provided by two Atlas 300i NPU cards, features portability and silent design, and utilizes a customized liquid cooling system to achieve "extremely low noise" operation. This addresses the pain points of traditional AI devices, such as high noise levels and difficult deployment, making it suitable for noise-sensitive scenarios such as laboratories and offices. It has become the preferred device for AI localization applications in small and medium-sized enterprises (SMEs) ranging from 70 to 32b.
02
hardcore strength
Longjing Technology has established a clear "fully domestic" technology path, covering the entire stack of processors and achieving autonomy in GPU/NPU. It has achieved comprehensive computing power support from ARM to x86 architecture. The adoption of customized fans and liquid cooling systems in multiple products is not simply a matter of component assembly, but rather a deep technological integration and reinvention.
Fully-stack, autonomous and controllable chip architecture
Longjing has deep cooperation with chip manufacturers such as Feiteng and Kunpeng, optimizing BIOS and drivers for AI workloads.
Taking LH4402GN-8KA as an example, the processor affinity scheduling algorithm has increased the utilization rate of the computing power of the Ascend 910 NPU from 70% to 92%, resulting in a 31% increase in processing speed in image recognition tasks.
At the CPU level, we cover three major domestic chip systems: Feiteng (ARM architecture), Haiguang (x86 architecture), and Kunpeng (ARM architecture). In terms of acceleration chips, we adopt dedicated AI chips such as Ascend 910/NPU and Atlas 300i to achieve autonomous and controllable computing power supply.
Global application of liquid cooling technology
From data center models to portable workstations, Longjing has miniaturized and customized liquid cooling technology. Through structural optimization, customized fans, and liquid cooling technology, it has improved heat dissipation efficiency while achieving a 15% reduction in power consumption and an 80% reduction in noise, truly realizing "cool operation and green computing".
Server level: Increase the air intake area by 15% and enhance the heat dissipation efficiency by 50%;
At the workstation level: the volume is reduced by 40%, and the noise level is controlled below 45 decibels (equivalent to a library environment).
Hardware and software co-optimization
Longjing has implemented designs such as "tool-free maintenance", "out-of-band visual management", and "organic log automatic recording" across its entire product line, significantly reducing operational complexity. From hardware monitoring to fault warnings, from log analysis to remote maintenance, Longjing has established an intelligent service system that covers the entire product lifecycle, allowing customers to "buy with confidence and use with peace of mind".
The self-developed "Longzhi" management platform can monitor the health status of hardware in real-time.
Longjing's products are genuinely supporting the intelligent transformation of various industries in China.
03
Empowerment by large models
For enterprises, localization is merely the foundation; the key lies in the ability to rapidly generate value. The Linglong Digital Intelligence Cloud Platform (AIOC), jointly developed by Shanghai Longjing Technology and Shanghai Supercomputing Center, boasts three major modules: "Wenshu - Yanhai - Zhixi". Targeted at enterprise-level users, it truly addresses the pain point of the "last mile" in AI implementation.
Wenshu: An intelligent assistant deeply integrated with a knowledge base, supporting regulation inquiry and fault diagnosis, enabling efficient access to key information.
Yanhai: A powerful auxiliary tool that integrates intelligent writing, polishing, translation, and question-answering, significantly enhancing content creation efficiency.
ZhiXi: Under the premise of ensuring data security and privacy, it provides accurate document interpretation, intelligent question answering, and multidimensional knowledge guides.
The AIOC platform, based on a domestic AI computing framework, can operate stably on various domestic CPU servers and operating systems after deep optimization. It supports remote installation and local deployment, helping enterprises build a privatized AI service system that ensures "data remains within the domain and computing power is independently controlled", meeting the stringent requirements for data privacy and security compliance in key areas such as government affairs, finance, energy, and military industry.
Finally,
When the LD5112W-2HA workstation enters an ordinary office, and when the LG4215G-10FA supports the training of national-level AI large models,
These "Chinese cores" are redefining the boundaries of computing power.
Longjing Technology provides a secure, efficient, and green computing engine for various industries with its integrated solution of "hardware + software + services".
When we open the chassis of Longjing servers, what we see is not only a dense array of chips and circuits, but also the determination of Chinese technology enterprises to break through blockades. From the optimization of the instruction set of Feiteng processors to the fluid mechanics design of liquid cooling systems, these "invisible innovations" are reshaping the global computing industry landscape.
