In the document NVIDIA CUDA Toolkit v3.2 RC2 Release Notes for Windows
the following statement can be found:
“Vista, Server 2008 and Windows 7 related:
o In order to run CUDA on a non-TESLA GPU, either the Windows desktop must be extended onto the GPU, or the GPU must be selected as the PhysX GPU.”
I have computer with 3 GTX 480 GPUs. OS is Windows 7 x64 Professional. Monitor is connected to only one GPU, and in NVIDIA Control Panel (graphics driver 266.58 is installed) under 3D Settings/Set SLI and PhysX configuration, as PhysX processor, CPU is selected. However, I still can use all 3 GPUs to run CUDA programs on them.
Is there anything strange about it?