Are GPUs Non-Green Computing Devices?

• Martín Pi Puig ^[1] ; Laura De Giusti ^[1] ; Marcelo Naiouf ^[1]
  1. [1] Universidad Nacional de La Plata

     Universidad Nacional de La Plata

     Argentina

     
• Localización: Journal of Computer Science and Technology, ISSN-e 1666-6038, Vol. 18, Nº. 2, 2018, págs. 153-159
• Idioma: inglés
• DOI: 10.24215/16666038.18.e17
• Títulos paralelos:
  • ¿Son las GPUs dispositivos eficientes energéticamente?
• Enlaces
  • Texto completo
• Resumen
  • español

    Con el consumo de energía emergiendo como uno de los mayores problemas en el desarrollo de aplicaciones HPC (High Performance Computing), la importancia de trabajos específicos de investigación en este campo se convierte en una prioridad. En los últimos años, los coprocesadores GPU se han utilizado frecuentemente para acelerar muchos de estos costosos sistemas, a pesar de que incorporan millones de transistores en sus chips, lo que genera un aumento considerable en los requerimientos de energía. Este artículo analiza un conjunto de aplicaciones del benchmark Rodinia en términos de rendimiento y consumo de energía de CPU y GPU. Específicamente, se comparan las versiones secuenciales y multihilo en CPU con implementaciones GPU, caracterizando el tiempo de ejecución, la potencia real instantánea y el consumo promedio de energía, con el objetivo de probar la idea de que las GPU son dispositivos de baja eficiencia energética.

  • English

    With energy consumption emerging as one of the biggest issues in the development of HPC (High Performance Computing) applications, the importance of detailed power-related research works becomes a priority. In the last years, GPU coprocessors have been increasingly used to accelerate many of these high-priced systems even though they are embedding millions of transistors on their chips delivering an immediate increase on power consumption necessities. This paper analyzes a set of applications from the Rodinia benchmark suite in terms of CPU and GPU performance and energy consumption. Specifically, it compares single-threaded and multi-threaded CPU versions with GPU implementations, and characterize the execution time, true instant power and average energy consumption to test the idea that GPUs are power-hungry computing devices.

• Referencias bibliográficas
  • References [1] K. Kasichayanula, D. Terpstra. “Power Aware Computing on GPUs”. In: Symposium on Application Accelerators in High Performance...
  • [2] X. Mei, Q. Wang, X. Chu. A survey and measurement study of GPU DVFS on energy conservation. Digital Communications and Networks, vol....
  • [3] X. Ma, M. Dong, L. Zhong, Z. Deng. “Statistical Power Consumption Analysis and Modeling for GPU-based Computing”. In: HotPower ACM SOSP...
  • [4] H. Nagasaka, N. Maruyama. “Statistical Power Modeling of GPU Kernels Using Performance Counters”. In: International Conference on Green...
  • [5] R. Suda, Da Qi Ren. “Accurate Measurements and Precise Modeling of Power Dissipation of CUDA Kernels toward Power Optimized High Performance...
  • [6] S. Huang, S. Xiao, W. Feng. “On the energy efficiency of Graphic Processing Units for scientific computing”. In: IEEE Symposium on Parallel...
  • [7] J. Lang, G. Rünger. “High-Resolution Power Profiling of GPU Functions Using Low-Resolution Measurement”. In: Euro-Par 2013 Parallel Processing...
  • [8] NVML Reference Guide, http://docs.nvidia.com/deploy/nvml-api/index.html. Accessed on 2018-04-13.
  • [9] S. Che, M. Boyer, J. Meng, D. Tarjan, J. W. Sheaffer, S. Lee, K. Skadron. “Rodinia: A benchmark suite for heterogeneous computing”. In:...
  • [10]OpenMP Application Programming Interface, http://www.openmp.org/wpcontent/uploads/openmp-4.5.pdf. Accessed on 2018-04-13.
  • [11]10 K. Asanovic et al. “The landscape of parallel computing research: A view from Berkeley”. Technical Report UCB/EECS-2006-183, EECS Department,...
  • [12]S. Che, J. W. Sheaffer, M. Boyer, L. G. Szafaryn, L. Wang, K. Skadron. “A characterization of the Rodinia benchmark suite with comparison...
  • [13]Running Average Power Limit – RAPL, https://01.org/blogs/2014/running-average-power-limit-%E2%80%93-rapl. Accessed on 2018-04-13.
  • [14]13 E. Rotem, A. Naveh, D. Rajwan, A. Anathakrishnan, E. Weissmann, “Powermanagement architecture of the Intel microarchitecture codenamed...
  • [15]M. Burtscher, I. Zecena, Z. Zong. “Measuring GPU Power with the K20 Built-in Sensor”. In: GPGPU-7 Proceedings of Workshop on General Purpose...
  • [16]V. Weaver, M. Johnson, K. Kasichayanula, J. Ralph, P. Luszczek, D. Terpstra, S. Moore. “Measuring Energy and Power with PAPI”. In: 41st...
  • [17]D. Li, S. Byna, S. Chakradhar. “Energy-Aware Workload Consolidation on GPU”. In: 40th International Conference on Parallel Processing...
  • [18]Y. Zhang, Y. Hu, B. Li, L Peng. “Performance and power analysis of ATI GPU: A statistical approach”. In: 6th IEEE International Conference...
  • [19]Y. Abe, H. Sasaki, S. Kato, K. Inoue, M. Edahiro, M. Peres. “Power and performance characterization and modeling of GPUaccelerated systems”....
  • [20] J. Chen, B. Li, Y. Zhang, L. Peng, J. K. Peir. “Tree structured analysis on GPU power study”. In: 29th IEEE International Conference...
  • [21]GCC Command Options: Options That Control Optimization, https://gcc.gnu.org/onlinedocs/gcc-4.6.2/gcc/Optimize-Options.html. Accessed on...