Research on Collaborative Optimisation of High-power Chip Multi-dimensional Thermal Management Technology and Integrated Devices
Abstract
In response to the thermal management bottlenecks brought about by the high integration and high powerization of high-power chips, as well as the performance compatibility problems of on-chip integrated interconnection and energy storage devices, this paper carries out multi-structure coupled thermal management, network-on-chip (NoC) architecture fault tolerance optimisation, on-chip energy storage device material modification and high-performance thermal interface. Through numerical simulation and experimental verification, the performance laws of bionic microchannels jet shock-coupled heat dissipation, gradient porous-heat pipe composite heat dissipation, topology optimisation-microchannel integrated liquid cooling, and dry ice particle jet phase change heat dissipation are revealed; a machine learning-driven NoC fault tolerance core mapping model is built, and the packaging is supplemented with the channel shortening and equalisation technology of the line NoC. The research results show that the bionic coupled heat dissipation system can reduce the maximum temperature of the 800W chip by 8.47%, the topology optimisation microchannel integrated liquid cooling structure achieves 4.28K cooling, the dry ice jet system realises the safe temperature control of the chip at 300W/cm² heat flow density, and the liquid metal TIM at 90W Compared with the traditional hot silicone grease, the temperature is 9.8℃, and the RI-CSF scheme of wireless NoC achieves a significant SINR improvement, and the capacity of integrated energy storage devices is increased to 34.2 mF/cm². This research provides technical support for the collaborative design of high-efficiency thermal management and integrated devices of high-power chips.
Fields of Study & Subjects
Heat Transfer and Optimization
Parallel Computing and Optimization Techniques
Phase Change Materials Research
Publication details
- Journal / source
- Applied and Computational Engineering
- Publisher
- EWA Publishing
- Volume
- 220
- Issue
- 1
- Pages
- 138–146
- Publication date
- 2 February 2026