\titlesize Generative Design for Direct-to-Chip Liquid Cooling for Data Centers
Zheng Liu University of Michigan-Dearborn Dearborn, MI · 2026
Read \titlesize Generative Design for Direct-to-Chip Liquid Cooling for Data Centers by Zheng Liu University of Michigan-Dearborn Dearborn, MI (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2604-10941.txt.
Description
Liu presents a generative frame for direct-to-chip cooling channels for NVIDIA's GB200 Superchip, where rising AI workloads raise power densities and heuristic layouts miss non-uniform temperatures. A finite-difference model gives fast temperature predictions and feeds feedback to a reaction-diffusion process generating channel topologies under constraints. Iterating generation and evaluation shifts cooling toward high-power regions, curbing hot spots. Against baseline, the channels cut average temperature by over 5 degrees and maximum temperature by over 35 degrees. Generative design plus lightweight modeling can boost direct-to-chip cooling.
Catalog record
| CREATOR | Zheng Liu University of Michigan-Dearborn Dearborn, MI |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 13508 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2604.10941