Slashdot Mirror
Carmack Shows Off the id Tech 5 Engine
MojoKid writes "id's keynote
address from this past week's Quakecon featured John Carmack revealing details of the id tech 5 engine. For more on the subject, GamesIndustry.biz has an interview with id developer Steve Nix about the project. 'I know that when we started working with Splash Damage on Enemy Territory they wanted large, detailed outdoor terrains, and they had some ideas on how to dynamically load the textures and everything, and John [Carmack] said, 'Why don't we try this new approach and make the entire terrain one massive texture, and then just load blocks of texture in dynamically that you can see at any one given time?' So John did the initial work on it, got it up and running, and it just so happened that that work was the basis for what we have in id Tech 5.'"
Here's a few videos of Carmack talking about it:
- Part 1
- Part 2
- Part 3
RegardselFarto
The difference apparently is between dynamically loading a whole bunch of different textures (as-needed, which is the traditional way of doing things), and paging through 1 texture.
Previously, you had to break up your textures into blocks and manually load them as the player moved into the area covered by those textures. This new technology allows you to create one large texture for the entire level and have the game engine automatically break up the texture into blocks and dynamically load them for you.
It makes programming easier, because its one less thing to keep track of.
We all know what to do, but we don't know how to get re-elected once we have done it
It also opens up the possibility to stream these textures in line-by-line, instead of block-by-block. No need to waste time/memory loading in a 2048x2048 texture of the next area as soon as it's barely visible if we can do it arbitrarily.
That's because it is DirectX9, atleast on the XBox360, OpenGL on the PC/Mac and OpenGL|ES/whatever else they use on the PS3.
RegardselFarto
You're missing the point of the tech demonstrations. Depth of field, motion blurring, and HDR lighting are current-gen features and are commonly featured in games out today. He was rather focusing on the amount of detail (80 GB worth) that was dynamically loaded, and didn't rely on old "hacks" (detail textures heavily repeated over a lower-res diffuse map, etc) to create the illusion of detail at an acceptable frame rate. And do so while keeping the engine both portable and fast is icing on the cake.
That in itself is rather impressive. As an OpenGL developer, I'm no longer impressed by motion blur, volume shadows, and other textbook shaders/effects - I'm more impressed by this.
Sigs are for losers
Eesh! Repeat. id do NOT make the games you want. id make ENGINES and games that showcase them.
A little harsh maybe, but basically correct. If you want a rip-roaring game with a smashing storyline, wait until the first game companies start licensing the tech from id. Until then, like me, you could do a lot worse than forgetting your high-minded, high-falluting city-slicker ambitions and just run around for a while, shooting every demon between you and the blue keycard.
Sorry you're completely wrong. The mega texture itself is not based on tiles, that's just how they author it. It all gets baked out into a single completely unique texture at the end. So the initial large scale detail really is hugely detailed and extremely high res. Tiling is just a way of quickly painting terrain for outdoor areas, and then you can use an unlimited number of overlays (because they get baked out into the texture) to for example stamp out rocks and stuff to break up the surface. You can easily envision having a mode where you hit a key which brings a screenshot up into photoshop, you edit it however you want, and then project it back into the scene.
Shader results aren't really stored in video memory (unless you want them to be). They're output directly to the buffer based on a combination of shader code and base textures. Although the terrain may generate quite a bit of data, the resources it used to generate that frame won't be as large.
Eventually we'll all want video cards with 10GB of memory, but this is a nice way of bridging the gap.
Rage uses lightmapping on the environment. The only thing that really moved in the Rage trailer was the cars. The cars cast a shadow even if they are inside another shadow (from the same light source)