RTX 5090 COLD PLATE LIQUID-COOLED SERVER

Designed specifically for scenarios with explosive AI computing power demands, this high-density computing core leverages cold-plate liquid cooling technology and an 8-GPU cluster of RTX 5090 graphics cards. It meets the stringent PUE (Power Usage Effectiveness) requirements of China’s "Eastern Data and Western Computing" project, providing a stable computing power foundation for high-heat-density tasks such as large model training and supercomputing simulations. It stands as a benchmark model for the large-scale application of liquid cooling technology in 2025.
1. Cold-Plate Liquid Cooling: Leap in Heat Dissipation Efficiency
The copper cold plate fits tightly with the GPU/CPU cores. Combined with precise temperature control via a 16KW CDU (Cooling Distribution Unit), the graphics card core temperature is reduced by 15-20°C compared to air cooling, enabling continuous full-load operation for 8 hours in a 50°C environment.
The closed-loop liquid cooling circulation system adopts a fanless design, with an operating noise level below 55dB, making it suitable for quiet scenarios such as laboratories and medical facilities.
2. High Density & Stability: Sustained Computing Power Output
The 8-GPU direct connection architecture reduces data latency, supporting the training of dozens of industry-specific large models (e.g., DeepSeek-R1). The training efficiency for 10-billion-parameter models is 3 times higher than that of air-cooled servers.
Equipped with dual power supplies and a dual variable-frequency pump redundancy design, its MTBF (Mean Time Between Failures) exceeds 100,000 hours, meeting the 24/7 requirements of scientific research and production.

GPU REPAIR SERVICE

Jieyi Technology offers professional GPU and server maintenance services, covering scenarios like GPU card dropout, PCB board faults, nvswitch unrecognition, firmware update, and backplane faults. The standardized process involves users submitting fault descriptions, engineers conducting quick testing and quotation, professional teams performing maintenance after plan confirmation, and equipment being sent back or retrieved. It has successfully repaired cases such as personal RTX 4090 cards, university H100 servers, enterprise A100 servers, and data center H800 servers, ensuring efficient equipment recovery and supporting customers' business continuity.​

GPU COMPUTING POWER LEASING SERVICE 说明：

Gogpu platform under Jieyi Technology provides cost-effective GPU computing power leasing services, covering popular graphics cards like H100, 5090, 4090, A100, and A800, with flexible billing by hour/day/week/month. It offers three machine types (container, cloud host, bare metal), realizes GPU resource containerization and pooling via Hong Kong University of Science and Technology team's technology to ensure multi-tenant isolation. With multiple national intelligent computing centers, it serves over 50,000 individual users and 500+ SMEs, operating compliantly to support AI large model training, inference, etc., with elastic scaling.​