The same rule applies, when the world is three times as big, your team needs to 10X, unless you render lots of water, lots of desert...
So we were talking about how the new xbox has roughly 10 times the CPU power of the old one.
What does it really mean?
I was using an open world game as an example, say it has the draw distance of 1KM.
The open world game will have to render 360 degrees of the world, so it has to render an area of 1KM^2 X Pi = 3.14KM^2.
So if I I have to render an area with radius of 2KM, how much more CPU power do I need?
2KM^2 X Pi = 12.57KM^2 (stuff to render)
12.57/3.14 (2KM / 1KM) = 4
Wow, to double the distance, I will need 4X the CPU power!
So 10X CPU power gives you
sqr of 10 = 3.1623KM
That means, when we 10X the CPU power, the world is 3.16KM instead of 1KM in an open world game...
This refutes Ben's article - Endless open worlds
Of course this is pure maths, and there are plenty of interesting things you can do to increase perceived draw distance.
Ok, one last maths equation
if I increase 1KM to 100KM draw distance, how much more CPU power do I need [trollface]
100^2 * Pi = 31,415 KM to render
31,415 / 3.1415 = 10,000 X CPU Power - that's a lot of CPU power
written by Roy Hui - Find me on Google Plus
Desi said: And further to Doyl's comment, not only is there frustum culling in video graphics (stuff outside the camera doesn't get rendered), but there's also occlusion culling (stuff behind opaque stuff doesn't get rendered) and backface culling (only the visible sides of an object get rendered). There are other tricks too, like specifically disabling the rendering of areas which you know will not be visible from a particular spot, which is how the Battlefield games do their large-scale occlusion culling. Sure these require some resources to work out, but they save a lot more than they use.
In PC and other platforms that have dedicated GPUs, the CPUs get used for AI, physics calculations (if the GPU isn't being used for them), game state, and sundry visual effects.
The thing that limits draw distance the most, in my opinion, is system memory. You've gotta store not only the current and surrounding cells, but also the states of items and actors within those cells, and their relationships to each other. You've also got to recompute them per screen draw (or at least fairly regularly), and keep them updated in memory. More gigamoos of processing power can help, but all of that info must be stored, and it ain't getting stored in the hard disk, I tell you what. Too slow. It's a good thing that the new gen consoles are getting more RAM, finally the industry can use the RAM in our PCs properly.
(As an aside, "recompute per screen draw" is a big deal. That calculation get performed 60 or more times per second. That's why so many console devs chose to target 30 frames per second; more twice as much calculation time. I predict that the only tangible difference that will come from more CPU power is the ability to target something like 60fps or 90fps more reliably).