Simulation of the game world. How right?

Question

Game development

Interested in approaches to simulating the game world on the server.
Lyrics:
All read literature describes some kind of authoritarian server architecture.
All calculations are carried out on it. The client only sends commands.
So I'm wondering, in order to be able to determine if the avatar has hit the wall - you need to have on the server the entire geometry of the game world in the form of ... what?
Of the approaches known to me at the moment - geometry on primitives.
BUT. How to export all this from engines such as Unreal, Unity, ... I
also have my own idea on this matter - to keep an instance of the finished world, as on the client. But in this case, the server must have a video card. Thought, feel still possesses that genius, but still.
Can you point me in the right direction?

Answer 1

The server has a complete world map for client control.
On clients, the world map is partial, with loading from the server as needed - for a picture to the player.
But, on the client of the world with the objects needed for rendering.
And on the server, lightweight data structures, for example, BSP (or whatever in Unreal), deeply integrated into the world and used for visualization on the client, are not needed on the server.
Ideally, these are 2 completely differently organized maps / worlds. Synchronized.
In practice, to save time, developers can keep almost identical worlds. Although this creates a decent (unnecessary) load.

Answer 2

But in this case, the server must have a video card.

A video card is needed to display the image on the monitor. Even if you are
Video card on the server is still not needed.