Hi NVIDIA Team and Community,
I would like to share a foundational architecture I have been working on regarding extreme thermal management for large-scale AI clusters and high-density data centers: the MicroChannel-DirectFlow architecture.
This architecture innovatively combines silicon-based 3D micro-channel etching directly on the Bare Die or TSV gaps with adaptive variable-frequency micro-jetting. By utilizing an electroosmotic pump and phase-change fluid circulation, it achieves extreme heat transfer efficiency (> 5000 W/(m²·K)) and can drop AI core temperatures by more than 25°C, effectively targeting hotspot thermal resistance. It is designed specifically for Supercomputing, LLM Data Centers, and Automotive Autonomous Driving SoCs.
To help foster industry innovation, the physical structure, channel topology, and control logic of this foundational architecture have already been open-sourced to the global public domain under the Creative Commons CC0 1.0 Universal License.
Meanwhile, I have retained a set of entirely new, advanced core designs. These are currently being evaluated by my legal team to explore potential patent applications.
I am very much looking forward to the opportunity to collaborate with the NVIDIA team in the future. If anyone from the team or the community has a moment to share your valuable insights or feedback on this invention, I would be deeply grateful.
You can find the open-source whitepaper and technical details attached below:
Zenodo URL Link: 微流控芯片集成式超导热微通道液冷与动态热平衡架构 (MicroChannel-DirectFlow): Open-Source Whitepaper | Zenodo
Thank you!
LIEW HSIAO LING