Carmack and Megatexturing

[eagl]

Carmack is a god.  I'm a comp sci weenie so this fascinates me...  His observation that texture tiles is really just a specific application of data compression is spot-on, and he's the only one outside of my PhD comp sci graphics professor who's ever pointed that sort of thing out.  This article is pure comp sci godness, and I hope it finds an application here in AH.

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

[MotleyCH]

That probably works well for small maps. Quake Wars maps are small..1-2 square miles.

[eagl]

Carmack's point is that it is boundless.  You can put as much or as little effort into detailed texturing as you like, but it's all one large texture paged in a bit at a time as required instead of one little texture brought in all at once and tiled.

I think it would make the trees and airfield boundaries look a hell of a lot better.  Roads and trains could also be brought in at whatever level of detail the arena designer wanted.  They could "paint" the texture at any desired detail level for the entire arena.  Increase detail level where you expect players to spend time (airfields, GV choke points, etc), and you could still apply generic terrains everywhere else.  The only thing you lose (apparently) is having to deal with matching up texture boundaries.

[Saintaw]

(Question from the clueless) But... wouldn't working on one huge tile make it a pain in the *** whenever you have to modify only a part of it ?(say, for a patch... you would have to re-download the whole 'terrain' tile).

IIRC (My memory fails sometimes... so, my appologies if i am incorrect), HTC uses a 3rd party 3D engine... so we would have to wait for that company to come up with this... then HTC gets AH3 coded in two weeks etc...

I keep reminding myself what games looked like 10 years ago compared to now, and I was already amazed back then.

[eagl]

You're right, it *could* result in a rather large terrain texture, but the idea is that it's infinately scalable up to where you hit whatever stop matters the most to you.

Plus, generic compression techniques ought to still work.  The big point is that texture sizes won't be limited to one "tile" of fixed size...  The developer can pick a "tile" size that is as big or small as they want.

[ChickenHawk]

Wow, good article and great technology.  The impact Carmack has had on the industry so far is immeasurable and this latest technique could potentially have a huge impact.

From an artistic standpoint, there are now no limits with this technology and no requirement for repetition anymore.  It will be more about how much time the developers want to give the artists to complete a map as they could potentially keep working on it for years on a large map, trying to make everything unique.

Of course this technology is still in it's infancy and I don't think HTC will be implementing it any time soon as Carmack isn't going to just give it away.  And the team at HTC would have to start from scratch and take away valuable development time from CT or new planes.  But it would be cool to have a unique environment.

Coincidentally, I just last week loaded up the original Unreal on a lark and installed the final patch that allows you to render using gforce and OpenGL cards and up the resolution.  I have to say, for a game that is around eight years old, I was blown away by how good the graphics really were.  Carmack has always been ahead of his time and he really is a game god.

[Chairboy]

Regarding the point about having to download the entire map, that's not true.  Through delta-based file updates, you can create a patch that only sends the difference between the original and updated version.  When I worked at Symantec, we did that all the time.  It kept the patches really small for the code.  They didn't do that with the virus defs because there were so many out there, but for code... it was perfect.