Emerging heterogeneous MPSoCs present significant challenges for accurate and efficient thermal modeling and management, as non-uniform thermal distributions become increasingly severe and cannot be neglected. In our latest TPDS paper, we address this issue by proposing 3D-ICE 3.1, an open-source non-uniform thermal simulator. You can access it here: https://lnkd.in/eSeG75Xj
Based on non-uniform 3D-ICE 3.1, an accelerated thermal evaluation framework is designed to support the swift development of advanced Dynamic Thermal Management (DTM) strategies. As a case study, we demonstrate how multi-agent reinforcement learning (MARL) can be leveraged to optimize the thermal, performance, and energy efficiency of heterogeneous 3D MPSoCs.
If you are interested, check the paper for more information: https://lnkd.in/emvs-vJR
Thanks to Luis M. Costero and David Atienza!
Darong Huang, Luis M. Costero and David Atienza are cooling down hot spots, with 3D-ICE!
This new publication explores the topic of dynamic thermal management for 3D configurations of microchips, using the 3D-ICE emulator developed at our lab.
Published this month in IEEE Transactions on Parallel and Distributed Systems:
https://lnkd.in/emvs-vJR
--
1 周Nice photo ?? b