Topology Optimization of MIMO Cooling Plates for Discrete Heat Sources in GPUs
Modelling · 14.06.2026 · 7 (3) · 116
DOI: 10.3390/modelling7030116 ↗Kurz gefasst
Fachartikel in der Zeitschrift Modelling, online erschienen am 14.06.2026. Verfasst von Jinzhao Fan und 4 weiteren Autor:innen. Die Arbeit nennt 47 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).
Abstract (Originalsprache)
With the rising integration of high-performance GPUs, localized hotspots induced by discrete heat sources present severe thermal challenges. Traditional single-inlet–single-outlet liquid cold plates can scarcely meet the heat dissipation requirements of inhomogeneous high heat fluxes. This study systematically investigates the effects of nine multiple-inlet–multiple-outlet (MIMO) configurations, ranging from single-inlet–single-outlet to three-inlet–three-outlet, on cold plate hydrothermal performance. An innovative stepwise optimization strategy, topology optimization (TO)-driven channel layout combined with fin-enhancement (FE)-based fine regulation, is proposed and verified to precisely regulate surface temperature distribution of discrete heat sources. The results show that the three-inlet–three-outlet configuration C-3 exhibits the optimal comprehensive performance among the nine configurations. Compared with the worst configuration A-2, C-3 reduces the pressure drop by 58.37% to only 147.18 Pa and yields the highest PEC, striking the optimum trade-off between heat transfer enhancement and fluid flow resistance. Through multi-inlet flow distribution and multi-outlet heat extraction, C-3 accurately suppresses heat accumulation in high heat flux regions, limiting the maximum temperature to merely 29.82 °C and drastically narrowing the substrate temperature difference from 8.69 °C to 2.12 °C. In comparison with the traditional cold plate (TCP), the optimized cold plate (OCP) realizes a 17.42% increase in performance evaluation criterion (PEC). Furthermore, the fin-enhanced optimized cold plate (FEOCP) reduces the temperature standard deviation by 54.15% relative to TCP, significantly enhancing temperature uniformity with only an additional pressure drop penalty of 5.43%. This study reveals the regulation mechanism of MIMO configurations on the flow field distribution of liquid cold plates and verifies the effectiveness of the TO-FE optimization framework, thus providing highly valuable engineering solutions for the high-efficiency, uniform-temperature and low-resistance heat dissipation of high-power electronic devices.
Zitieren
APA
Fan, J., Zhang, B., Liu, J., Cai, Y., & Shi, H. (2026). Topology Optimization of MIMO Cooling Plates for Discrete Heat Sources in GPUs. Modelling, 7(3), 116. https://doi.org/10.3390/modelling7030116
Vancouver
Fan J, Zhang B, Liu J, Cai Y, Shi H. Topology Optimization of MIMO Cooling Plates for Discrete Heat Sources in GPUs. Modelling. 2026;7(3):116. doi:10.3390/modelling7030116
BibTeX
@article{Fan2026,
author = {Fan, Jinzhao and Zhang, Bixiao and Liu, Jiazhen and Cai, Yufei and Shi, Hong},
title = {Topology Optimization of MIMO Cooling Plates for Discrete Heat Sources in GPUs},
journal = {Modelling},
year = {2026},
volume = {7},
number = {3},
pages = {116},
doi = {10.3390/modelling7030116}
}
EUVIMED