IEEE Micro · 2021 · 20 citations · 6 references
EngineeringComputer EngineeringElectrical EngineeringHeat TransferMicroelectronicsThermal AnalysisDevice ModelingHeat Transfer AnalysisHeat Transfer ModelingThermal DesignThermal ManagementThermoelectric MaterialsEnergy HarvestingEnergy Harvesting SystemsMulticore ArchitecturesThermal Modeling Of Thin FilmsThermoelectric CoolingThermoelectric Energy HarvestingThermoelectric Thin FilmsThin Film Thermoelectrics
On-demand cooling is inevitable to maximize the processor's performance, while fulfilling thermal constraints-this holds more in advanced technologies, where localized hotspots change during runtime. In this work, we propose to adopt an array of thin-film thermoelectric (TE) devices, which is integrated within the chip packaging, for both cooling and energy-harvesting purposes. Each TE device within the array can be during the runtime enabled either for energy harvesting or on-demand cooling. Our approach is implemented and evaluated using a mature finite elements analysis tool in which a commercial multicore mobile chip is modeled after careful calibrations together with state-of-the-art TE devices. Results demonstrate that our approach reduces the peak temperature by up to 24 <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">°</sup> C°C and the average temperature by 10 <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">°</sup> C°C across various benchmarks with the cost of 67.5 mW. Additionally, the harvested energy from the array of TE devices compensates for 89
6