GreenLLM – huge reductions in GPU power consumption
Congratulations to the authors of GreenLLM, boasting 34% reduction in total energy consumption compared to the default DVFS baseline, with no loss of throughput, and less than 3.5% increase in SLO violations. GreenLLM: SLO-Aware Dynamic Frequency Scaling for Energy-Efficient LLM ServingLiu, Qunyou; Huang, Darong; Zapater, Marina; Atienza, David20262026 63rd ACM/IEEE […]
New publication on system-accelerator co-design
Congratulations to Qunyou, Marina and David: Mitigating the Bandwidth Wall via Data-Streaming System-Accelerator Co-Design Liu, Qunyou; Zapater, Marina; Atienza, David 2026-08 ACM Transactions on Architecture and Code Optimization
3D-ICE Version 4.0 is out!
3D-ICE 4.0 is a powerful 3D Interlayer Cooling Emulator, which can deliver accurate and efficient thermal modeling for modern 2.5D/3D heterogeneous chiplet systems. 3D-ICE 4.0 supports modeling detailed material distribution across layers, directly imported from industrial layouts. To handle complex multi-layer stacks, 3D-ICE introduces adaptive layer partitioning and thermal-aware non-uniform […]
3D-ICE cooling – 20 year retrospective published
Congratulations to David Atienza, Kai Zhu, Darong Huang and Luis Costero for this excellent paper, published in IEEE Design & Test. This survey covers more than two decades of research on power and thermal modeling and management in modern processors. The authors compare analytical, regression-based, and neural network-based techniques for […]