TY - GEN
T1 - Energy Efficient Frequency Scaling on GPUs in Heterogeneous HPC Systems
AU - Kraljic, Karlo
AU - Kerger, Daniel
AU - Schulz, Martin
N1 - Publisher Copyright:
© 2022, The Author(s), under exclusive license to Springer Nature Switzerland AG.
PY - 2022
Y1 - 2022
N2 - With most major corporations and research institutions having pledged to support sustainability goals for High Performance Computing (HPC), energy efficiency is a critical factor when evaluating heterogeneous HPC systems. However, many popular hardware performance & energy measurement frameworks, such as LIKWID, and benchmarks, such as the STREAM or the hipBone benchmark, do not or not fully support execution on heterogeneous systems containing AMD or NVIDIA Graphical Processing Units (GPUs), leading to a gap with regards to the understanding the relationship between frequency, performance and energy. We aim at closing this gap by extending the performance measurement framework LIKWID to support both AMD and NVIDIA GPUs. We run the STREAM and hipBone benchmark on AMD and NVIDIA GPUs at different GPU core frequencies. We show that the minimum period between two measurements for our GPU is at least 100ms and that GPUs have a sweet spot with regards to energy consumption at approximately 75% of their maximum frequency with energy savings up to 30% at a performance overhead between 0.72% and 3.12%.
AB - With most major corporations and research institutions having pledged to support sustainability goals for High Performance Computing (HPC), energy efficiency is a critical factor when evaluating heterogeneous HPC systems. However, many popular hardware performance & energy measurement frameworks, such as LIKWID, and benchmarks, such as the STREAM or the hipBone benchmark, do not or not fully support execution on heterogeneous systems containing AMD or NVIDIA Graphical Processing Units (GPUs), leading to a gap with regards to the understanding the relationship between frequency, performance and energy. We aim at closing this gap by extending the performance measurement framework LIKWID to support both AMD and NVIDIA GPUs. We run the STREAM and hipBone benchmark on AMD and NVIDIA GPUs at different GPU core frequencies. We show that the minimum period between two measurements for our GPU is at least 100ms and that GPUs have a sweet spot with regards to energy consumption at approximately 75% of their maximum frequency with energy savings up to 30% at a performance overhead between 0.72% and 3.12%.
KW - Energy awareness
KW - Heterogeneous computing
KW - High Performance Computing
UR - http://www.scopus.com/inward/record.url?scp=85144814451&partnerID=8YFLogxK
U2 - 10.1007/978-3-031-21867-5_1
DO - 10.1007/978-3-031-21867-5_1
M3 - Conference contribution
AN - SCOPUS:85144814451
SN - 9783031218668
T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
SP - 3
EP - 16
BT - Architecture of Computing Systems - 35th International Conference, ARCS 2022, Proceedings
A2 - Schulz, Martin
A2 - Trinitis, Carsten
A2 - Papadopoulou, Nikela
A2 - Pionteck, Thilo
PB - Springer Science and Business Media Deutschland GmbH
T2 - 35th International Conference on Architecture of Computing Systems, ARCS 2022
Y2 - 13 September 2022 through 15 September 2022
ER -
