CUDA via rpm versus nvidia drivers via run file

I installed CUDA 10.0 using the rpm repo package procedure on NVDIA’s page following the installation guide. I thought that was supposed to install the nvidia drivers? After I did that the 4 GeForce RTX 2080 Ti (there is also motherboard video) cards were not using any driver. I rectified that by downloading and installing the most recent driver from NVIDIA for those cards via a runfile. I know years ago you had to install nvidia drivers first before CUDA but I thought that had changed. Is this truly the proper procedure? I don’t want to end up with a mix that will cause issues down the road.

Thanks for verifying!