What happens to onboard video when installing a CUDA GPU?

I have a small-form factor Lenovo 10GT002SUS which has Intel onboard video. It has a low-profile PCIe 3.0 x16 slot and two x1 slots. I would like to add a low-profile GPU to x16 slot.

What happens to the onboard video when the GPU card is added? Will that video be disabled and rerouted through the GPU? Can the GPU service two monitors and CUDA simultaneously? (The onboard video services two monitors.)

What’s a good low-profile GPU to get for a machine like this?It will be used to run Matlab programs that take more than a week to run.

The behavior is system-specific. There is no general answer.

Some systems allow the onboard video to remain in use. Both the onboard video and external GPU can be active.
Some systems automatically disable the onboard video when they detect a GPU plugged into a slot.
Some systems have BIOS selectable options to control behavior.

I’m sure Lenovo can recommend GPUs that they have tested in the system you mention.

If the integrated video were to be disabled, would the CUDA GPU work both as a graphics card and a CUDA GPU independently?

In general, yes, you can run CUDA code and display graphics just fine on a single NVIDIA card.

However, if your system is able to boot from the integrated video WHILE you have a GPU plugged in, that might be the ideal setup in the case that you are running CUDA kernels that take a long time to execute, which might bog down your screen refresh rate if you were to run (demanding) CUDA code and displaying graphics to the screen at the same time on the NVIDIA GPU.