NVIDIA Nsight in VSCode much slower than cuda-gdb

Wondering if any one has the similar problem.
I created a binary that runs some simple calculation in CUDA. Setting the same breakpoint in VSCode it takes significantly longer to reach than if i run it in cuda-gdb. As far as I know the Nsight also uses cuda-gdb so this is very strange.

And this has happened consistently no matter what kernel code I write, and this has happened through out different re-install of my computer, so this is a consistent symptom as far as I am aware.

Wondering if any one has similar experience or insight, greatly appreciate it.

To make sure this is tracked properly, is this a comparison with VS Code and Visual Studio and cuda-gdb? If Visual Studio is not involved, I will move this to the Nsight VSCode Edition forum. Nsight Visual Studio Code Edition - NVIDIA Developer Forums