A closed-loop generative framework using finite-difference thermal modeling and constrained reaction-diffusion produces cooling channels for the NVIDIA GB200 that cut average chip temperature by more than 5°C and maximum temperature by over 35°C versus a standard parallel-channel baseline.
Generative design and optimiza- tion of battery packs with active immersion cooling,
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
eess.SY 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Generative Design for Direct-to-Chip Liquid Cooling for Data Centers
A closed-loop generative framework using finite-difference thermal modeling and constrained reaction-diffusion produces cooling channels for the NVIDIA GB200 that cut average chip temperature by more than 5°C and maximum temperature by over 35°C versus a standard parallel-channel baseline.