A dGPU with separate DRAM memory can be connected to the Tegra device over PCIe or NVLink. It is currently supported only on the NVIDIA DRIVE platform
If you use AGX Orin developer kit, coneecting to dGPU is not supported. Please note this.
For allocating memory on iGPU, if you would like to allocate memory which ca be accessed through GPu an dCPU, you can call cudaMallocManaged(): CUDA Runtime API :: CUDA Toolkit Documentation