First call to 'nppiResize_8u_C4R' takes ~20 seconds on Ubuntu 20.04 without sudo

Accelerated Computing CUDA CUDA Programming and Performance
4

Adding some important additional information.

Recreated the issue on 4 different laptops, all running Ubuntu 20.04, which have the following GPUs:

1. NVIDIA GeForce RTX 3050 Ti Laptop GPU
2. NVIDIA GeForce RTX 3050 Ti Laptop GPU
3. NVIDIA GeForce RTX 3050 Ti Laptop GPU
4. NVIDIA RTX A4500 Laptop GPU

I found another Ubuntu 20.04 on which the issue does not happen (i.e my application works perfectly without sudo).
The working PC has an older GPU:

NVIDIA GeForce GTX 1650

All of these PCs have the same driver/cuda/OpenGL versions:

Driver Version: 470.182.03   
CUDA Version: 11.4
OpenGL version string: 4.6.0 NVIDIA 470.182.03

@Robert_Crovella do you have a suggestion to anything else I should compare between the “good” & “bad” PCs?