On a 3-way system I always see one data transfer per frame between GPUs in NSight DirectX Transfers -window. All I do is take a texture from DXGI swap chain, clear it, then present. Is this normal or could there be something wrong with my setup of DXGI or device?
Related topics
| Topic | Replies | Views | Activity | |
|---|---|---|---|---|
| GPU to GPU transfers most effective method? | 27 | 38264 | March 3, 2011 | |
| Real-time GPU processing Peer 2 peer data copy, Linux kernel memory, kernels in kernel, | 35 | 8192 | June 30, 2010 | |
| Is there a way to transfer data between 2 GPU DIRECTLY? | 3 | 3914 | October 13, 2010 | |
| DMA transfers in parallel 2-way SLI with 2 GTX 280 | 8 | 3742 | March 16, 2009 | |
| PCI Transfer directly to GPU anytime soon? | 8 | 3762 | November 30, 2010 | |
| Nsight 3.1 with VS2012 DX11 WITHOUT swap-chain (cmd line renderer) | 1 | 1225 | August 27, 2013 | |
| Transfer data from PCIe device to GPU memory? Why not incorporate the SLI bridge into CUDA? | 15 | 11599 | March 28, 2010 | |
| Using GTX 590 cards for CUDA SLI cards under CUDA? | 37 | 14326 | April 2, 2012 | |
| Pixel transfer is synchronized with 3D rendering - Bug? | 0 | 3357 | December 18, 2012 | |
| CUDA vs DX execution times DX GPGPU code --> CUDA = slower | 15 | 13336 | January 30, 2008 |