Modeling and Chasing the Energy-Efficiency Sweet Spots in Modern GPUs
2026-07-01 • Distributed, Parallel, and Cluster Computing
Distributed, Parallel, and Cluster Computing
AI summaryⓘ
The authors studied how different hardware settings affect the energy used to complete tasks on computers with CPUs and GPUs. They ran tests using molecular simulation programs and a stress test on various GPUs and CPUs, changing their speed and power limits. They found that energy use changes depending on the specific task and hardware, with certain speeds making the system less efficient. The authors created a simple model to explain how GPU power use changes with speed, helping to find the best settings to save energy. Their work shows that adjusting GPU speed is usually better for saving energy than just limiting power.
Energy-to-solutionCPU-GPU systemsMolecular dynamicsFrequency scalingPower cappingGPU power modelingThermal design powerEnergy efficiencyHeterogeneous computingWorkload characterization
Authors
Ayesha Afzal, Markus Manfred Li, Michael Panzlaff
Abstract
Energy consumption is a key limitation in high-performance computing on heterogeneous CPU-GPU systems. This work studies how hardware configuration affects energy-to-solution under realistic workloads. We study energy efficiency regimes using molecular dynamics benchmarks (GROMACS and AMBER) and a stress-test benchmark (FIRESTARTER) on systems with A40, A100, H100, and H200 GPUs and Intel Ice Lake CPU, varying frequency scaling and power cap. We show that energy-to-solution exhibits workload- and architecture-dependent transitions between efficient and inefficient regimes, driven by nonlinear GPU power-frequency scaling. We introduce an interpretable analytical model that decomposes GPU power into linear and nonlinear components, identifying a workload- and architecture-dependent transition frequency beyond which efficiency degrades. The model fits empirical data with low error and highlights the role of baseline power, nonlinear power behavior, and transition frequency as the dominant parameters governing energy efficiency. Power capping is generally less effective for efficiency tuning than frequency reduction, especially for workloads that operate far from thermal design power. Overall, energy-efficient HPC execution is a configuration-dependent problem with identifiable regime shifts, and we provide model-driven guidance for selecting operating points.