Designed for high-performance supercomputers and AI large-model computing servers, this supercomputer liquid cold plate adopts precision microchannel structures to attach directly to high-power chips such as GPUs and CPUs. Circulating coolant removes high-density heat flux from chips, lowering junction temperature. It enables ultra-high computing density per rack and helps optimize the PUE of data centers.

 

 

The supercomputer liquid cold plate, also known as HPC microchannel cold plate, is a core thermal component for high-performance computers and AI computing servers, with an application field entirely different from fiber laser cold plates.

 

As the computing power of GPUs and CPUs keeps rising, chip heat flux surges sharply, making traditional air cooling inadequate. This cold plate closely contacts the chip surface with dense micron-scale microchannels fabricated internally. Coolant flows through channels to directly extract heat from chips. Manufactured by vacuum brazing, friction stir welding (FSW) or diffusion bonding, the cooling cavity delivers high sealing performance and low thermal resistance. It achieves uniform temperature control on chip surfaces and eliminates local hotspots.

 

Custom cold plates are available for standard HGX GPU modules, CPUs, VRMs and memory modules. Copper or aluminum alloy material options are provided, compatible with deionized water and propylene glycol coolant. Widely deployed in national supercomputing centers, AI computing clusters and research-oriented high-performance computing platforms.

 

 

High-Efficiency Microchannel Heat Transfer & Low Thermal Resistance Precision internal microchannel arrays greatly expand heat exchange area, delivering far stronger cooling capacity than conventional cold plates. It withstands ultra-high chip heat flux and effectively reduces chip junction temperature.

 

Excellent Surface Temperature Uniformity, Hotspot Elimination CFD fluid simulation optimizes flow channels for even flow distribution and small temperature variation across the chip. It prevents thermal throttling and unstable computing performance caused by local chip overheating.

 

High-Strength Sealed Cavity for Safety & Reliability Fabricated by vacuum brazing, diffusion bonding or FSW friction stir welding. Every unit undergoes hydrostatic pressure test and helium mass spectrometer leak detection before shipment. Low leakage risk protects high-value computing chips.

 

Flexible Material Selection for Diverse Working Conditions Copper is preferred for high-power variants for superior thermal conductivity; aluminum alloy is selected for lightweight versions. Passivation or nickel plating surface treatment provides corrosion resistance and compatibility with multiple coolant types.

 

Modular Customization Compatible with Mainstream Computing Modules Customizable geometry for HGX, various CPUs, VRMs and PCIe switch modules. Mounting holes, positioning posts and quick connector positions can be tailored to match rack liquid cooling systems in servers.

 

Low Flow Resistance Design to Reduce CDU Load Optimized channel structure controls pressure drop while maintaining cooling capacity, cutting power consumption of Cooling Distribution Units (CDU) and reducing overall power usage of data centers.

 

 

Contact Substrate High-precision flat substrate with tightly controlled flatness and roughness. Used together with thermal pads or thermal grease to closely mate with GPU/CPU chips and minimize interfacial thermal resistance.

 

Microchannel Heat Exchange Core Dense micron-scale flow channels enabling sufficient heat exchange between coolant and metal, serving as the core heat transfer zone of the cold plate.

 

Manifold Chamber Diversion and collection structure ensuring even coolant distribution into every microchannel to avoid flow maldistribution.

 

Fluid Inlet & Outlet Ports Standard industrial quick-disconnect or threaded connectors optional for fast plug-and-play maintenance and connection to Cooling Distribution Units (CDU).

 

Mounting & Positioning Structure Locating pins and locking mounting holes ensure precise alignment between cold plate and chips with consistent assembly quality.

 

 

✅ National supercomputing centers and high-performance HPC clusters ✅ AI large-model training/inference computing servers, complete HGX GPU modules ✅ High-density cloud computing servers ✅ Research computing platforms and semiconductor test equipment ✅ Liquid cooling retrofitting projects for green data centers