Title :
A comparative benchmarking of the FFT on Fermi and Evergreen GPUs
Author :
Ahmed, Mohamed F. ; Haridy, Omar
Author_Institution :
Comput. Sci. & Eng., American Univ. in Cairo, Cairo, Egypt
Abstract :
NVIDIA and AMD GPUs are are gaining traction in HPC for their performance and architectural aspects. It is very important to measure and analyze the relative power of each architecture. In this paper, we analyze the architecture of NVIDIA´s Fermi and AMD´s Evergreen processors and demonstrate the best practices and techniques to best utilize the capabilities of each architecture. We implemented the FFT on both cards utilizing our findings to reach new performance ceilings on both GPUs.
Keywords :
benchmark testing; computer graphic equipment; coprocessors; fast Fourier transforms; performance evaluation; storage management; AMD Evergreen GPU; FFT; NVIDIA Fermi GPU; architecture analysis; comparative benchmarking; performance ceiling; Bandwidth; Concurrent computing; Graphics processing unit; High definition video; Instruction sets; Pipelines;
Conference_Titel :
Performance Analysis of Systems and Software (ISPASS), 2011 IEEE International Symposium on
Conference_Location :
Austin, TX
Print_ISBN :
978-1-61284-367-4
Electronic_ISBN :
978-1-61284-368-1
DOI :
10.1109/ISPASS.2011.5762726
