How big is a Continuum world?
Continuum is built to hold a universe at 1:1 scale. Here is what that means in practice, with a few real numbers from the engine.
No render-distance limit
There is no render-distance limit and there are no loading screens. You can look from a room across a city, or from orbit across a planet, and the world keeps going.
Storage grows with edits
Holding a universe would be hopeless if every voxel had to be stored separately. In Continuum, storage grows linearly with the number of edits you make. Repeated structure can share storage, and each copy can still be edited on its own.
Four quadrillion cubes from twelve pieces
One test world is a Menger sponge, the fractal you get by splitting a cube into 27 smaller cubes, keeping 20 of them, and repeating the same step inside each one. Taken to 12 levels, it has 531,441 of the smallest cubes along each edge and 4,096,000,000,000,000 of them in total. That is about four quadrillion.
Before any edits, the engine builds it from 12 unique construction nodes. Those four quadrillion cubes are represented, not stored one by one, and you never see all of them at once. The structure repeats inside itself all the way down.
A million copies of one street
Another test repeats the Amazon Lumberyard Bistro, a well-known street scene released through NVIDIA's Open Research Content Archive (ORCA) under CC BY 4.0, in a grid of 1,024 by 1,024. That is 1,048,576 copies. Each copy keeps its own address in the world, so you can change one without changing the rest.
The copies sit in a grid with gaps between them, so this is a million separate scenes rather than one continuous city. The Bistro is Amazon's work, not ours.
What these numbers do not say
None of these numbers say anything about frame rate on a particular machine. They describe how much world the engine can represent and edit.