Êtes-vous un étudiant de l'EPFL à la recherche d'un projet de semestre?
Travaillez avec nous sur des projets en science des données et en visualisation, et déployez votre projet sous forme d'application sur GraphSearch.
Modern high-performance processors employ thermal management systems, which rely on accurate readings of on-die thermal sensors. Systematic tools for analysis and determination of best allocation and placement of thermal sensors is therefore a highly relevant problem. Moreover liquid cooling has emerged as a promising solution for addressing the elevated temperatures in 3D Multi-Processor Systems-on-Chips (MPSoCs). In this work, we present a combined sensor placement and convex optimization approach for thermal management in 3D-MPSoC with liquid cooling. This approach first finds the best locations inside the 3D-MPSoC where thermal sensors can be placed using a greedy approach. Then, the temperature sensing information is subsequently used by our convex-based thermal management policy to optimize the performance of the MPSoC while guaranteeing a reliable working condition. We perform experiments on a 3D multicore architecture case-study using benchmarks ranging from web-accessing to playing multimedia. Our results show a reduction up to 10× in the number of required sensors. Moreover our policy satisfies performance requirements, while reducing cooling energy by up to 72% compared with traditional state of the art liquid cooling techniques. The proposed policy also keeps the thermal profile up to 18 C source lower compared with state of the art 3D thermal management techniques using variable-flow liquid cooling.
Sophia Haussener, Felix N. Büchi, Mirco Magnini
David Andrew Barry, Ulrich Lemmin, Naifu Peng, François Mettra, Rafael Sebastian Reiss