Beyond DVFS: A first look at performance under a hardware-enforced power bound

Barry Rountree, Dong H. Ahn, Bronis R. De Supinski, David K Lowenthal, Martin Schulz

Research output: Chapter in Book/Report/Conference proceedingConference contribution

100 Citations (Scopus)

Abstract

Dynamic Voltage Frequency Scaling (DVFS) has been the tool of choice for balancing power and performance in high-performance computing (HPC). With the introduction of Intel's Sandy Bridge family of processors, researchers now have a far more attractive option: user-specified, dynamic, hardware-enforced processor power bounds. In this paper we provide a first look at this technology in the HPC environment and detail both the opportunities and potential pitfalls of using this technique to control processor power. As part of this evaluation we measure power and performance for single-processor instances of several of the NAS Parallel Benchmarks. Additionally, we focus on the behavior of a single benchmark, MG, under several different power bounds. We quantify the well-known manufacturing variation in processor power efficiency and show that, in the absence of a power bound, this variation has no correlation to performance. We then show that execution under a power bound translates this variation in efficiency into variation in performance.

Original languageEnglish (US)
Title of host publicationProceedings of the 2012 IEEE 26th International Parallel and Distributed Processing Symposium Workshops, IPDPSW 2012
Pages947-953
Number of pages7
DOIs
StatePublished - 2012
Event2012 IEEE 26th International Parallel and Distributed Processing Symposium Workshops, IPDPSW 2012 - Shanghai, China
Duration: May 21 2012May 25 2012

Other

Other2012 IEEE 26th International Parallel and Distributed Processing Symposium Workshops, IPDPSW 2012
CountryChina
CityShanghai
Period5/21/125/25/12

Fingerprint

Hardware
Electric potential

Keywords

  • Power bound
  • RAPL

ASJC Scopus subject areas

  • Software

Cite this

Rountree, B., Ahn, D. H., De Supinski, B. R., Lowenthal, D. K., & Schulz, M. (2012). Beyond DVFS: A first look at performance under a hardware-enforced power bound. In Proceedings of the 2012 IEEE 26th International Parallel and Distributed Processing Symposium Workshops, IPDPSW 2012 (pp. 947-953). [6270741] https://doi.org/10.1109/IPDPSW.2012.116

Beyond DVFS : A first look at performance under a hardware-enforced power bound. / Rountree, Barry; Ahn, Dong H.; De Supinski, Bronis R.; Lowenthal, David K; Schulz, Martin.

Proceedings of the 2012 IEEE 26th International Parallel and Distributed Processing Symposium Workshops, IPDPSW 2012. 2012. p. 947-953 6270741.

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Rountree, B, Ahn, DH, De Supinski, BR, Lowenthal, DK & Schulz, M 2012, Beyond DVFS: A first look at performance under a hardware-enforced power bound. in Proceedings of the 2012 IEEE 26th International Parallel and Distributed Processing Symposium Workshops, IPDPSW 2012., 6270741, pp. 947-953, 2012 IEEE 26th International Parallel and Distributed Processing Symposium Workshops, IPDPSW 2012, Shanghai, China, 5/21/12. https://doi.org/10.1109/IPDPSW.2012.116
Rountree B, Ahn DH, De Supinski BR, Lowenthal DK, Schulz M. Beyond DVFS: A first look at performance under a hardware-enforced power bound. In Proceedings of the 2012 IEEE 26th International Parallel and Distributed Processing Symposium Workshops, IPDPSW 2012. 2012. p. 947-953. 6270741 https://doi.org/10.1109/IPDPSW.2012.116
Rountree, Barry ; Ahn, Dong H. ; De Supinski, Bronis R. ; Lowenthal, David K ; Schulz, Martin. / Beyond DVFS : A first look at performance under a hardware-enforced power bound. Proceedings of the 2012 IEEE 26th International Parallel and Distributed Processing Symposium Workshops, IPDPSW 2012. 2012. pp. 947-953
@inproceedings{d8c12365991d42d48cc2d6ff24c02e50,
title = "Beyond DVFS: A first look at performance under a hardware-enforced power bound",
abstract = "Dynamic Voltage Frequency Scaling (DVFS) has been the tool of choice for balancing power and performance in high-performance computing (HPC). With the introduction of Intel's Sandy Bridge family of processors, researchers now have a far more attractive option: user-specified, dynamic, hardware-enforced processor power bounds. In this paper we provide a first look at this technology in the HPC environment and detail both the opportunities and potential pitfalls of using this technique to control processor power. As part of this evaluation we measure power and performance for single-processor instances of several of the NAS Parallel Benchmarks. Additionally, we focus on the behavior of a single benchmark, MG, under several different power bounds. We quantify the well-known manufacturing variation in processor power efficiency and show that, in the absence of a power bound, this variation has no correlation to performance. We then show that execution under a power bound translates this variation in efficiency into variation in performance.",
keywords = "Power bound, RAPL",
author = "Barry Rountree and Ahn, {Dong H.} and {De Supinski}, {Bronis R.} and Lowenthal, {David K} and Martin Schulz",
year = "2012",
doi = "10.1109/IPDPSW.2012.116",
language = "English (US)",
isbn = "9780769546766",
pages = "947--953",
booktitle = "Proceedings of the 2012 IEEE 26th International Parallel and Distributed Processing Symposium Workshops, IPDPSW 2012",

}

TY - GEN

T1 - Beyond DVFS

T2 - A first look at performance under a hardware-enforced power bound

AU - Rountree, Barry

AU - Ahn, Dong H.

AU - De Supinski, Bronis R.

AU - Lowenthal, David K

AU - Schulz, Martin

PY - 2012

Y1 - 2012

N2 - Dynamic Voltage Frequency Scaling (DVFS) has been the tool of choice for balancing power and performance in high-performance computing (HPC). With the introduction of Intel's Sandy Bridge family of processors, researchers now have a far more attractive option: user-specified, dynamic, hardware-enforced processor power bounds. In this paper we provide a first look at this technology in the HPC environment and detail both the opportunities and potential pitfalls of using this technique to control processor power. As part of this evaluation we measure power and performance for single-processor instances of several of the NAS Parallel Benchmarks. Additionally, we focus on the behavior of a single benchmark, MG, under several different power bounds. We quantify the well-known manufacturing variation in processor power efficiency and show that, in the absence of a power bound, this variation has no correlation to performance. We then show that execution under a power bound translates this variation in efficiency into variation in performance.

AB - Dynamic Voltage Frequency Scaling (DVFS) has been the tool of choice for balancing power and performance in high-performance computing (HPC). With the introduction of Intel's Sandy Bridge family of processors, researchers now have a far more attractive option: user-specified, dynamic, hardware-enforced processor power bounds. In this paper we provide a first look at this technology in the HPC environment and detail both the opportunities and potential pitfalls of using this technique to control processor power. As part of this evaluation we measure power and performance for single-processor instances of several of the NAS Parallel Benchmarks. Additionally, we focus on the behavior of a single benchmark, MG, under several different power bounds. We quantify the well-known manufacturing variation in processor power efficiency and show that, in the absence of a power bound, this variation has no correlation to performance. We then show that execution under a power bound translates this variation in efficiency into variation in performance.

KW - Power bound

KW - RAPL

UR - http://www.scopus.com/inward/record.url?scp=84867417218&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84867417218&partnerID=8YFLogxK

U2 - 10.1109/IPDPSW.2012.116

DO - 10.1109/IPDPSW.2012.116

M3 - Conference contribution

AN - SCOPUS:84867417218

SN - 9780769546766

SP - 947

EP - 953

BT - Proceedings of the 2012 IEEE 26th International Parallel and Distributed Processing Symposium Workshops, IPDPSW 2012

ER -