Carmack on MegaTexturing

[Section_Ei8ht]

Found this Q&A with Carmack talking about MegaTexturing, his latest thing that he did for Quake Wars.

http://www.gamerwithin.com/?view=articl ... 1319&cat=2

From my understanding MegaTexturing is an efficient way to put a 32000x32000 texture on a terrain without the need for 4 gigs of video ram.

Sounds like right now its only being applied to terrain, but I think there will soon be an implementation that uses a texture for all geometry in the map, so there is no repetition of textures. But then again, maybe that would be kinda pointless because no one would notice and it would just be a waste of time for a developer to make a 300ft long brick wall that is not a repeating texture.

Don't know. Sounds cool. What do you all think?

[von*ferret]

interesting read.

[csharp]

thats amazing.. And great technology But i think that its a bit hard to make 32000*32000 sized texture

[Hourences]

Texture tiling isnt that bad if you do it correctly so imo there are more important things to improve than this. I also wonder who is going to be able of making a 32000 or even a 8192*8192 texture. Youll need lots of ram and lots of time. I dont think thats very efficient.

The filesize will also be huge.

Still if it works its nice and its always good to have as many methods supported as possible so you can choose the thing that suits your specific situation best but to use this technique all over the place? I dont see that happening right now.

[Pericolos0]

when i look at the quake wars screenshots i really see no difference with other terrain based engines. I can see how the technology is useful though, but it doesnt seem like an essential breakthrough. Just another buzzword

[D3ads]

Does anyone really stare at the ground long enough to give two shits about how sharp a 32000x32000 texture is? I mean really? No, we'll be too busy shooting Strogg in the face and vice versa.

Yet more unecessary graphical features that increase development time :roll:

[Thrik]

Read the article next time D3.

Anyway, it sounds like a fairly good idea, since it's not like you can't still just duplicate a texture in certain parts if you don't feel like creating every square inch.

Having all the terrain as one big texture is good from a development standpoint though as you can easily move bits around, assuming you have stuff on layers. It makes more sense to me to have everything on one master texture since you can easily get a good idea of how it all fits together, and don't have to keep track of loads of seperate files.

It's certainly no breakthrough, but it is a nice way of doings things and could help to lessen development time depending on the project needs. Obviously it's no good for a lot of types of game, especially those with many indoor areas.

[Polaris]

Does anyone really stare at the ground long enough to give two shits about how sharp a 32000x32000 texture is? I mean really? No, we'll be too busy shooting Strogg in the face and vice versa.

Yet more unecessary graphical features that increase development time :roll:

Q3: Aside from the visual aspect of the terrain looking better, do you think there will be any other foreseeable differences to us gamers that are playing MegaTexture games?

Answer: It’s just the variety and the diversity of it. Like I said at the very beginning, this is only a very small aspect of graphics, let alone of games in the larger sense. It’s a specific little piece of technology that addresses texture resource limitations, and this entire technology would not need to exist if you had four gigabyte graphics cards, and lots more RAM

You have to take a step forward somewhere, even if its only something small like this. I doubt it increases dev time by much though.

[TomWithTheWeather]

I wonder if the artist creates it or the level designer creates it...

[Zacker]

I don't really think that this matters a lot. A properly made texture with the right size and scale does not give any tiling issues today. So far the ground textures in QW does not look impressive at all.

Hourences, I think the point of the system would be that it would not require a lot of ram. As I read it it, the system would automaticly split it up and load the pieces in a dynamic way.

[Defrag]

It's not just for terrain. In the article, Carmack states that in the latest revisions, he has extended the system for general purpose use. It sounds really interesting to me; it gives artists the flexibility to have basically unlimited unique resources (if they so choose) without RAM hogging. I'm unsure as to whether it'll have a huge impact like Carmack claims, but he does make some interesting points (namely that, for all our technological advances, we're still using heavily 'compressed' schemes for our data [small, tiling textures]). As long as this fits into the game asset creation pipeline(s) in an elegant fashion, it could prove to be very cool.

[Polaris]

it'll have a huge impact like Carmack claims

When did he say that?

[Section_Ei8ht]

it'll have a huge impact like Carmack claims

When did he say that?

he didn't

Answer: Well for the user the bottom line is just that it looks better. You wind up with something that has the diversity that you don’t get with more conventional terrain generation systems out there. As the developer, looks are still important for games. If you look at a game and you make it look better, it’s a better game, so long as you don’t impact the gameplay negatively. So it’s nothing profound and fundamental, it’s just one tiny little aspect of graphics rendering that’s just better now.

As long as this fits into the game asset creation pipeline(s) in an elegant fashion, it could prove to be very cool.

Don't think it will be that much of a problem. Asset creation with id tech has always been a breeze. I'm sure anyone who has worked with the Doom 3 engine can back me up on that.

[Hourences]

Hourences, I think the point of the system would be that it would not require a lot of ram. As I read it it, the system would automaticly split it up and load the pieces in a dynamic way.

Im sure he can code his engine to do that as described in the article but I doubt Adobe are going to code in the same support for Photoshop.

Im not going to open and edit 32000*32000 images with multiple layers in PS..

[von*ferret]

I dont suppose photoshop can reference files like Max has "xref" files?