Performance Analysis of Fermi GPU Architecture

초록

NVIDIA introduced a GPU (Graphic Processing Unit) architecture called G80 for GPGPU (General Purpose computing on GPU) in 2006; however, performance is not good because of delay time for accessing off-chip memory. To complement this defect, Fermi architecture was introduced in 2009. In this paper, we analyze the difference between G80 and Fermi architecture, and propose the implementation method.

키워드

GPGPUFermi ArchitectureGrobal memoryBinary search
제목
Performance Analysis of Fermi GPU Architecture
저자
Lim, Ji HyukKim, Dong Kyue
발행일
2011-07
유형
Proceeding
저널명
Conference Proceedings MITA 2011
MITA2011 Proceeding
페이지
146 ~ 147