Thought this was interesting. Says posted 19 hours ago. http://www.gamerwithin.com/?view=article&article=1319&cat=2

Quote: Q6: So would you consider the fact that the MegaTexture paints all of the terrain with one enormous texture an advantage to level of detail or a limitation? Answer: Level of detail wise, the terrain does not render with any sophisticated geometry morphing situation. That’s one of those things that for years I think most of the research that’s gone into has been wasted. Geometry level of detail on terrain…there have been thousands of papers written about it, and I honestly don’t think it’s all that important. The way the hardware works, you’re so much better off setting down a static mesh that’s all in vertex and index buffers, and just letting the hardware plow through it, rather than going through and having the CPU attempt to do some really clever cross blended interpolation of vertices. In and infinite sized world, you would have to include some degree of level of detail. The Quake Wars levels are not infinite size. They’re bounded. And it generally turns out to be the best idea to just have the geometry at a reasonable level of detail and very efficiently rendered. But the MegaTexture would work just fine if you wanted to use that on something where you were dynamically level detailing the terrain. That is one of the nice aspects of it, where to the application it just looks like you can texture with an infinite size texture. You don’t have to worry about breaking it up on particular boundaries of anything special like that.

NVMD

[Edited by - Smit on December 4, 2006 12:46:25 AM]

Quote:
just let them treat one uniform geometry mesh and have this effectively unbounded texture side on there, and use a more complicated fragment program to go ahead and pick out exactly what should be on there, just as if the graphics hardware and the system really did support such a huge texture.

Im confused as to what they are actually talking about. Reading this sentence over and over makes me think that they have one fairly large texture like 1024x1024 where each pixel color represents a proceedural texturing method for an area. I mean obviously there is finite memory on the card. Can anyone clarify what this technology actually is? Right now it seems like a some buzzword to sell a video game.

Looking at the screenshots, the megatextures (the ground) are not even noticeable to me. In some shots they actually look more low resolution then current tiled solutions. For example there are some random paper on the ground of one shot with a bunch of really generically bumpmapped looking dirt, that almost looks like a brown-colored stretched noise texture. All the screenshots look like they could be made with tiled textures. What am I missing?

He goes on to talk about how everyone has been caught up on per-pixel lighting, shadowing effects, ect.. He so surprised no one has dived into "megatextures" before, but I have yet to see one screen of something, where I go that is one really big non-repeating texture... wow. At least when I look at a screen shot of bumpmapping I can notice the difference between that and a vertex lit model. I have a feeling that it is just a way of defining various proceedural texturing methods to various areas. Correct me if Im wrong.

Perhaps the texture is just in systemmem but because you only access it locally it is only cached in VRAM locally.

nkm

well even in system memory would a 32K texture be to large. So I presume they are using a very good compression algorithm for there texture, such as JPEG 2000, which even support 'regions of interests'. You can then update the texture every few frames, and do the decompression of the JPEG on the CPU. The problem with this solution is of course the use/abuse of the CPU for decompression and the upload of the texture every few frames. Decompression on the GPU would also be possible of course, and that would explain the fragment cost, but it would use more than 8 MB of video memory then.

With heavy compression system, I wouldn't know what they are doing on the fragment shader, but I presume he simply puts another few detail maps on top of it, and let the tiling be dependend on the distance from the camera. I tried to add this to our engine and this works rather nice. Only rather expensive on the GPU if you want to use upto 4 normal maps to blend in between but it works :)

Regards,
Kenzo