I just tested this against the Daz version in an extreme test. I only rendered for 50 iterations both without any denoising, and with Daz's denoiser and then applied the NVidia denoiser and here are my results. I'm using PNGs so there's no JPEG artifacts introduced.
First, the original image, 50 iterations, no denoising...
View attachment 2449647
Now, the same image, 50 iterations with Daz denoising set to 50, so it only denoises the final iteration...
View attachment 2449651
And finally, the external NVidia denoiser applied to the first noisy image...
View attachment 2449653
My own observations is there is very little difference between the two, at least in this example. BUT, I did notice one problem, and that is with her hair transparency around her eyebrows. With the Daz denoised version you can see all of her eye and eye brows, even behind the hair. but the Nvidia, external denoiser hides the eye and eyebrow that are covered by her hair.
I'll stick with Daz's denoiser. The external one is a nice one, especially if you're using an AMD GPU, but the DAZ one still seems superior, or at least the same to me.
I think the hair covering her eyes problem could be due to how Daz has a Post Denoiser Alpha option. Which is a plus.