Quantitative analysis of CPU/GPU Co-execution in high-performance computing systems

SeungGu Kang, HongJun Choi, JaeHyung Park, Sung Woo Jung, JongMyon Kim, DongSeop Kwon, JoongChae Na, Cheol Hong Kim

Research output: Contribution to journalArticle

Abstract

GPUs have been widely used in most off-the-shelf high-performance computing systems, since CPUs could not meet the increasing throughput demands efficiently. Therefore, the performance of up-to-date high-performance computing systems can be maximized when the task scheduling between the CPU and the GPU is optimized. In this paper, we analyze CPU and GPU co-execution in the perspective of performance, energy efficiency and temperature, depending on task scheduling. Usually, GPU execution leads to better performance and better energy efficiency than CPU execution when single application is executed. However, in cases that multiple applications are executed, GPU execution cannot guarantee better performance and better energy efficiency than CPU execution, depending on application characteristics. Especially, the system behavior becomes more unpredictable when multimedia applications are executed compared to the cases that computation-intensive applications are executed. We also analyze the performance, energy efficiency and temperature of computing systems varying the GPU types. Experimental results show that high-end GPUs provide better performance and energy efficiency than low-end GPUs, while the temperature of high-end GPUs goes higher than mat of low-end GPUs.

Original languageEnglish
Pages (from-to)2923-2936
Number of pages14
JournalInformation
Volume15
Issue number7
Publication statusPublished - 2012 Jul 1

Fingerprint

Program processors
Chemical analysis
Energy efficiency
Scheduling
Graphics processing unit
Temperature
Throughput

Keywords

  • CPU
  • CUDA
  • GPU
  • High-performance computing
  • Scheduling

ASJC Scopus subject areas

  • General

Cite this

Kang, S., Choi, H., Park, J., Jung, S. W., Kim, J., Kwon, D., ... Kim, C. H. (2012). Quantitative analysis of CPU/GPU Co-execution in high-performance computing systems. Information, 15(7), 2923-2936.

Quantitative analysis of CPU/GPU Co-execution in high-performance computing systems. / Kang, SeungGu; Choi, HongJun; Park, JaeHyung; Jung, Sung Woo; Kim, JongMyon; Kwon, DongSeop; Na, JoongChae; Kim, Cheol Hong.

In: Information, Vol. 15, No. 7, 01.07.2012, p. 2923-2936.

Research output: Contribution to journalArticle

Kang, S, Choi, H, Park, J, Jung, SW, Kim, J, Kwon, D, Na, J & Kim, CH 2012, 'Quantitative analysis of CPU/GPU Co-execution in high-performance computing systems', Information, vol. 15, no. 7, pp. 2923-2936.
Kang S, Choi H, Park J, Jung SW, Kim J, Kwon D et al. Quantitative analysis of CPU/GPU Co-execution in high-performance computing systems. Information. 2012 Jul 1;15(7):2923-2936.
Kang, SeungGu ; Choi, HongJun ; Park, JaeHyung ; Jung, Sung Woo ; Kim, JongMyon ; Kwon, DongSeop ; Na, JoongChae ; Kim, Cheol Hong. / Quantitative analysis of CPU/GPU Co-execution in high-performance computing systems. In: Information. 2012 ; Vol. 15, No. 7. pp. 2923-2936.
@article{51344c836d7740c98466d370df489bd8,
title = "Quantitative analysis of CPU/GPU Co-execution in high-performance computing systems",
abstract = "GPUs have been widely used in most off-the-shelf high-performance computing systems, since CPUs could not meet the increasing throughput demands efficiently. Therefore, the performance of up-to-date high-performance computing systems can be maximized when the task scheduling between the CPU and the GPU is optimized. In this paper, we analyze CPU and GPU co-execution in the perspective of performance, energy efficiency and temperature, depending on task scheduling. Usually, GPU execution leads to better performance and better energy efficiency than CPU execution when single application is executed. However, in cases that multiple applications are executed, GPU execution cannot guarantee better performance and better energy efficiency than CPU execution, depending on application characteristics. Especially, the system behavior becomes more unpredictable when multimedia applications are executed compared to the cases that computation-intensive applications are executed. We also analyze the performance, energy efficiency and temperature of computing systems varying the GPU types. Experimental results show that high-end GPUs provide better performance and energy efficiency than low-end GPUs, while the temperature of high-end GPUs goes higher than mat of low-end GPUs.",
keywords = "CPU, CUDA, GPU, High-performance computing, Scheduling",
author = "SeungGu Kang and HongJun Choi and JaeHyung Park and Jung, {Sung Woo} and JongMyon Kim and DongSeop Kwon and JoongChae Na and Kim, {Cheol Hong}",
year = "2012",
month = "7",
day = "1",
language = "English",
volume = "15",
pages = "2923--2936",
journal = "Information (Japan)",
issn = "1343-4500",
publisher = "International Information Institute",
number = "7",

}

TY - JOUR

T1 - Quantitative analysis of CPU/GPU Co-execution in high-performance computing systems

AU - Kang, SeungGu

AU - Choi, HongJun

AU - Park, JaeHyung

AU - Jung, Sung Woo

AU - Kim, JongMyon

AU - Kwon, DongSeop

AU - Na, JoongChae

AU - Kim, Cheol Hong

PY - 2012/7/1

Y1 - 2012/7/1

N2 - GPUs have been widely used in most off-the-shelf high-performance computing systems, since CPUs could not meet the increasing throughput demands efficiently. Therefore, the performance of up-to-date high-performance computing systems can be maximized when the task scheduling between the CPU and the GPU is optimized. In this paper, we analyze CPU and GPU co-execution in the perspective of performance, energy efficiency and temperature, depending on task scheduling. Usually, GPU execution leads to better performance and better energy efficiency than CPU execution when single application is executed. However, in cases that multiple applications are executed, GPU execution cannot guarantee better performance and better energy efficiency than CPU execution, depending on application characteristics. Especially, the system behavior becomes more unpredictable when multimedia applications are executed compared to the cases that computation-intensive applications are executed. We also analyze the performance, energy efficiency and temperature of computing systems varying the GPU types. Experimental results show that high-end GPUs provide better performance and energy efficiency than low-end GPUs, while the temperature of high-end GPUs goes higher than mat of low-end GPUs.

AB - GPUs have been widely used in most off-the-shelf high-performance computing systems, since CPUs could not meet the increasing throughput demands efficiently. Therefore, the performance of up-to-date high-performance computing systems can be maximized when the task scheduling between the CPU and the GPU is optimized. In this paper, we analyze CPU and GPU co-execution in the perspective of performance, energy efficiency and temperature, depending on task scheduling. Usually, GPU execution leads to better performance and better energy efficiency than CPU execution when single application is executed. However, in cases that multiple applications are executed, GPU execution cannot guarantee better performance and better energy efficiency than CPU execution, depending on application characteristics. Especially, the system behavior becomes more unpredictable when multimedia applications are executed compared to the cases that computation-intensive applications are executed. We also analyze the performance, energy efficiency and temperature of computing systems varying the GPU types. Experimental results show that high-end GPUs provide better performance and energy efficiency than low-end GPUs, while the temperature of high-end GPUs goes higher than mat of low-end GPUs.

KW - CPU

KW - CUDA

KW - GPU

KW - High-performance computing

KW - Scheduling

UR - http://www.scopus.com/inward/record.url?scp=84864326836&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84864326836&partnerID=8YFLogxK

M3 - Article

AN - SCOPUS:84864326836

VL - 15

SP - 2923

EP - 2936

JO - Information (Japan)

JF - Information (Japan)

SN - 1343-4500

IS - 7

ER -