Peer Device

Accelerated Computing GPU-Accelerated Libraries
1

I have a GTX 690 (2 device) and Im trying to access the peer:

int p1fromp0, p0fromp1;
HANDLE_ERROR(cudaDeviceCanAccessPeer(&p1fromp0, 0, 1));
HANDLE_ERROR(cudaDeviceCanAccessPeer(&p0fromp1, 1, 0));

However, the variable p1fromp0 and p0fromp1 has response equal 0. I dont understanding.

The GTX 690 dont has direct access of peerDevice from device?

Thanks,

2

Have you tried the simpleP2P CUDA SDK sample?

Have you checked that the GPUs are on the same I/O hub? This is a necessary condition to achieve peer-to-peer access?