MegaTexture/Visibility

[MrRee]

I’m wondering if some of you good people to maybe clear up something thats been bugging me since the whole MegaTexture announcement.

Having mapped in all iterations of the Quake engine, it’s always been of paramount importance to maintain strict control of the scene visibility to ensure top performance. This was enabled with good use of hint brushes and PVS, and more recently in D3 engines, well placed visportals.

With the huge draw distances expected from MegaTexture, what techniques need applied to achieve maximum performance?

[Ciabs]

I don’t know anything about mapping, but being the megatexture a unique massive texture that’s gonna be loaded anyway at map loading, I think there shouldn’t be any visibility/performance issues, since the map will be always fully loaded and always being using the same amount of VRAM, regardless if you’re in a narrow corridor or out in the open…

[kamikazee]

Actually they don’t use VIS at all for the terrain if you can believe Carmack. Here’s a quote Paul “Locki” Wedgwood, taken from this interview:

As John Carmack said in his keynote speech, we have a 32,000 by 32,000 unique texture that covers the entire terrain, so visually you have a unique landscape. You can have rocks, cliffs, and tarmac blending into sand, grass or gravel. And we can also draw all the way to the horizon.

You know how other developers will fog a big terrain so they can include more stuff? We don’t do that. We Brute Force Render right to the horizon as far as you can see. You can stand at one end of the map and look across it diagonally and see for a mile and a half.

[MuffinMan]

there still will be some vis control - it’s not only the floor texture after all, it’s trees, buildings, items, vehicles, players, explosions… I can’t imagine them to render everything and switch completely to box maps