CUDA: Specifying the graphics card on Macbook Pro 15 Retina

Questions
, , , ,
1

Hello,
This is a general question about running CUDA on a Mac (10.9.5) with retina display.

The newer Macbook Pros have two graphics cards. The FrontEnd and CUDA utilize the same higher performance graphics card.

Does anyone know if I can make the FrontEnd use the lower performance card?
Would this improve performance for my CUDA computations?
If both the FrontEnd and CUDA are in use, is there a risk of GPU memory being overwritten? I seem to get more frequent crashes when CUDALink is loaded.

Showing that CUDA is using the Nvidia card:

Cuda_Info.png
Cuda_Info.png701×253 51.2 KB

activity monitor showing FrontEnd used

Showing that FrontEnd is using the same card

activity_monitor.png

2

A developer suggested the third-party app from https://gfx.io/ .

3

Thanks Bruce,
That app works and shows that both the FrontEnd and a cuda program both can access the GPU at the same time without crashing the FrontEnd.

The question remains whether I can specify that the FrontEnd use a specific graphics card.

Many thanks, Craig