倍精度で計測
$ /home/t_azu/NVIDIA_GPU_Computing_SDK/C/bin/linux/release/nbody -benchmark -fp64 Run "nbody -benchmark [-n=<numBodies>]" to measure perfomance. -fullscreen (run n-body simulation in fullscreen mode) -fp64 (use double precision floating point values for simulation) > Windowed mode > Simulation data stored in video memory > Double precision floating point simulation > Compute 2.1 CUDA device: [GeForce GT 430] 2048 bodies, total time for 10 iterations: 80.215 ms = 0.523 billion interactions per second = 15.687 double-precision GFLOP/s at 30 flops per interaction
約16GFLOPS。理論性能では単精度:倍精度は12:1との話だったと思いますが、実測ではほぼ6:1。