After reading this i did a more in deep dive into nanite.
I tested a few meshes and how this all really works (Spoiler: I still dont understand how this works full).
So to me it looks like if you activate Nanite this will then also use some sort of decimation like in Zbrush too (you can see this also in the pictures).
Beside the size of the model it really seems like it doesn´t matter how many polygons it have.
The big benefit i see in using a lot of polys is not using a normal map. I tested 8k textures they where around 80mb and the normal map for this was around 150mb, ORM was 30mb. A model out of the tests i made was only 31,5mb compressed nanite size.
But i still try to figure out how to use the best out of both.
Maybe it would be a good idea to include a lower model because of the disk size it uses and also a normal map .
So everybody can decide himself what to use. And if you want to edit the Uv´s yourself it can be really hard with high res meshes if you dont have a really high end pc.
I also think it would be hard to build a very low poly model out of a model (with baked normals) that is build with Nanite in mind if you dont think about this from the start. And some things just need extra polys to look almost the same.
Below i added a few pictures from a few tests i made.
Here are a few tombstones both are Nanite. One is 500k and the other 100k without normal.
I guess you cant really tell which one is the one with 100k.
But the disk space used doesn´t really was a difference. 6,4mb full model – and the funny thing 7,7mb the decimated model.
So in this case it seems it would be better ro go with the higher res model.
But the very high model gives a complete different result
Base 1,5 million Verts - 31,5mb disk space : Decimated to 300k Verts - 6,8mb disk space.
In this case it makes a hugh difference and it would make sence to use the lower res model. If there is not something with Nanite i dont understand or see right now.
Maybe someone else have tested this and can give a better answer to this topic.