. Author manuscript; available in PMC: 2015 Nov 19.

Published in final edited form as: IEEE Conf High Perform Extreme Comput. 2014 Sep;2014:10.1109/HPEC.2014.7040981. doi: 10.1109/HPEC.2014.7040981

Table 4. Fraction of peak single precision FLOPS achieved for 3D FFTs. Technology details same as in Table 2.

Processor	Model	Year	Library	Utilization
CPU	Nehalem	2009	FFTW	5.1%
GPU	Tesla	2009	cuFFT	2.6%
CPU	Sandy Bridge	2014	MKL	53.9%
GPU	Fermi	2014	cuFFT	17.0%
GPU	Kepler	2014	cuFFT	7.6%