Small performance on nVidia Geforce GTX 480

Accelerated Computing CUDA CUDA Programming and Performance
1

Hello, All.

Few days ago I`ve installed 2 nodes with 4 nVidia Geforce GTX 480.

OS: CentOS 5.3

CPU: Intel Core-Quad

The hosts are connected through Mellanox Infiniband host.

Device driver has 260.19.21 version.

When I`ve tried to start MPI example from SDK(simpleMPI) it gives not appropriate results in comparison with 8800 GT.

nProcs         GTX 480            8800 GT

1             2.046746            0.7711229

2             4.038924            0.8121631

3             6.053905            0.9436841

4             8.149040            0.9960486

I`ve tried to test the cuda performance without mpi with the help of sdk example(matriMul) and it gives appropriate results.

2

Hello, All.

Few days ago I`ve installed 2 nodes with 4 nVidia Geforce GTX 480.

OS: CentOS 5.3

CPU: Intel Core-Quad

The hosts are connected through Mellanox Infiniband host.

Device driver has 260.19.21 version.

When I`ve tried to start MPI example from SDK(simpleMPI) it gives not appropriate results in comparison with 8800 GT.

nProcs         GTX 480            8800 GT

1             2.046746            0.7711229

2             4.038924            0.8121631

3             6.053905            0.9436841

4             8.149040            0.9960486

I`ve tried to test the cuda performance without mpi with the help of sdk example(matriMul) and it gives appropriate results.