A wave of new academic and government research is tackling one of computing's most stubborn problems: keeping advanced chips from cooking themselves. Semiconductor Engineering has highlighted a cluster of technical papers that, together, show how heat is shaping the next generation of chip design.
At the smallest scales, researchers from San Jose State University and Sandia National Laboratories examined "self-heating" in 3nm GAA-FET transistors used for SRAM memory, according to Semiconductor Engineering, studying how different substrate isolation techniques affect both heat buildup and resistance to radiation.
Heat is also a central concern as chips move light instead of electricity. A team from the University of Florida, ficonTEC Service, and Colorado State University proposed an AI-driven framework to map and manage thermal behavior in 3D integrated photonic circuits, which the paper notes are advancing high-performance computing.
Other groups are pushing photonics and electronics closer together. Researchers from MITRE, the University of Colorado Boulder, Sandia, the University of Arizona, and MIT reported a "fully monolithic" integration of piezo-optomechanical photonics with CMOS electronics. Separately, teams including the Centre for Quantum Technologies at the National University of Singapore described a room-temperature, CMOS-compatible photonic quantum processor.
The roundup also spans memory and automotive computing: Georgia Institute of Technology presented an "Open DRAM Model" for analyzing processing-in-memory in 3D DRAM, while the University of Stuttgart evaluated hardware abstraction layer concepts for software-defined vehicles.
Why it matters: as manufacturers pack more transistors into tighter 3D stacks and blend electronics with light, managing heat is becoming as important to performance and reliability as raw speed.