Like everything I write, this was initially... pages. So here's the condensed version.

Computer graphics. In all modern 3D games, they consist of texturing and lighting wireframes. Wireframes, things like these. Arrangements of triangles which represent solid surfaces.

In the past 10+ years, a lot of things have changed. In texturing, we used to just have one texture per object to give it the right color and pattern. Over time we've added bump mapping to simulate complex shadows caused by bumps and dips on otherwise flat surfaces, parallax mapping to let us see into those virtual dips and cracks, and specular and environmental mapping to tell the game how shiny surfaces are and what they should reflect. That's a big part of how we got from Half Life to Halo 3. So too have there been breakthroughs in lighting: where we once could only afford to have static lighting, we added volumetric shadows which show the shape of objects casting them, High Dynamic Range to make really bright objects seem brighter and better reflect how light interacts with our eyes, and even indirect lighting which allows light hitting objects to illuminate nearby objects--one of the reasons you never get a perfectly black shadow outside.

But you know what hasn't changed much? Those wireframes. Whether it's Mech Warior 2 or Cry Engine 3, our characters and powerups and landscapes are all ultimately the same wireframes, just varying in complexity. Sure, games will include multiple wireframes for everything these days, so you can see a zombie's individual fingers when they're up in your face but the computer doesn't waste time drawing every digit when they're a 100 yards off, but even that's a sloppy solution--what happens when you get really, really close to something, closer than the designers felt it worthwhile to add detail? Let's say you want to examine a trivial prop--maybe a faucet or a doorknob. In current games, the detail you see is limited to what the developers put in.

With any luck, that will soon change. With Windows 7 came the launch of DirectX 11. One of the features on that is hardware tessellation, which allows us to change and improve the way that we work with wireframes. Take a look at this. The model on the left was made by... I don't know, scanning somebody's head, or manually positioning the triangles. The model to the right of that, which is considerably more detailed, smooth, and realistic, was made entirely through an automated process (and the model all the way to the right was done with the same process applied a second time). Based on the 3D geometry of the original wireframe, the computer was able to divide each triangle into 4 smaller ones and calculate new positions for their corners to create a more curved and believable surface. It's the same stuff that Pixar does with their animated films to make sure characters don't have blocky noses and flat foreheads--they just make the triangles so tiny that you can't see any one individually. This is tessellation, and DirectX 11 brings support for it. This means that potentially, in a game which properly leverages this technology, you can get as close as you want to a curved surface and never see the flat shapes that make it up. Your computer just says "Uh oh, he's getting really close, we better make some more triangles!" and the object gets more detail. Obviously this won't be able to add extra details and imperfections to objects, like tire treads and bumps on somebody's skin, but your objects will be able to have smoother curves without the game developers wasting their time and storage space making a ten bajillion triangle wireframe for every pen cap and pebble in the game. Just imagine if your favorite games from a decade ago had this feature--you could play them today and while the textures might be blurry and the lighting simplistic, at least the character's hair, clothing, equipment, and the world around them would have reasonably smooth curves. It'd be much less painful to replay older games.

Of course, the other big plus for DirectX 11 adoption is that it means adopting the changes of DirectX 10 as well. A new shader model, more programable stuff you don't care about, geometery instancing, etc. The point is that it's updated and better, though with an adoption hampered by XP users stuck with DX9. With Windows 7 proving to be a pretty big hit (estimates are ~40 million units by the end of the year and another 140 million in 2010, but those are just estimates!), game developers will hopefully start developing for DX10/10.1/11 first and foremost... sooner rather than later. DirectX 9 was an exciting and evolutionary change in back in 2002, but we're coming up on its 7th birthday in about 3 weeks and I for one would be happy if it could retire before its 8th.

Questions, comments, always welcome. Or just geek out and share some cool video clips of pretty upcoming games. Also, please don't try to warp this into a "graphics vs gameplay" argument forum. That's so last year. A game should offer both. Finally, credit where credit is due:  kairhl initially and inadvertently prompted me to write this.