There is a thread in another forum in which a developer noticed a big difference in performance in a C++ AMP program when windows were opened and closed just prior to running his program ( http://social.msdn.microsoft.com/Forums/en-US/parallelcppnative/thread/d39b7b0d-8d2a-4eb3-b778-af3063b6a54e ). He noticed that he can improve the performance if he adjusts the power management mode for the GPU in the NVIDIA control panel (Windows) to prefer maximum performance. I wanted to see if that will work for me to see if it changes compute times for C++ AMP, CUDA and OpenCL programs, but I found that I can’t even open the NVIDIA control panel. As it turns out, my machine has three GPUs and the NVIDIA card that I want to set the defaults for does not have the display attached to it. That apparently prevents the control panel from appearing, and from setting the defaults for the NVIDIA GPU. The other two GPUs are AMD, one of which has the display. The NVIDIA GPU is accessible via CUDA, as well as C++ AMP and OpenCL. How can I set the power management mode for the NVIDIA GPU if the NVIDIA GPU does not have the display?