NGC GRAPHICS

Want to know about the GameCube's graphical abilities? Look no further.

Special Effects:

Fog

We all know what fog is. Fortunately, in GameCube games it may finally be used for atmosphere, rather than limiting the draw distance to save polygons. All fog is not created equally. You can bet that the GameCube's fog will look nicer than the N64's fog (which often manifested itself as a "wall" of fog rather than an atmospheric condition). Other kinds of fog don't always use this built in effect. These other fog-like effects might include smoke for example (see the Zelda demo).

Sub pixel anti-aliasing

This is a technique that is used to remove graphical defects known as "jaggies" and flickering from each frame. There are many known techniques for implementing anti-aliasing. Nintendo's SPAA is a modified form of super-sampling that provides a reasonble cross between performance and quality. The basic method behind SPAA is to draw the graphics at higher resolution (3x higher) and then blend the pixels down to normal TV resolution resulting in a better image. Check this article for more detailed information on anti-aliasing.

HW light x8

This feature can create accurate, dynamic lighting on a moving object that accurately represents 8 different lights sources per triangle. A great example of multiple light sourcing is Super Smash Bros. Melee. The trophy mode is a great place to start. You can change the lighting conditions with the press of a button as you rotate the trophy around. You should clearly be able to see teh effects of dynamic lighting. It's present in most other GameCube titltes too.

Alpha blending

This refers to hardware support for transparencies. This feature is used anytime you see glass, water or anything else that you can see through.

Virtual texture design

This is hardware support for moving textures around within the GameCube's memory architecture in small chunks. With this system, if you texture a giant billboard and only half of it is visible, the GameCube only fetches the visible parts of the texture, saving bandwidth and cache space. In addition, developers don't have to orchestrate texture swapping between memory banks. The "Flipper" handles it all automatically. What that basically amounts to is less hassle for game developers. That translates into happier, more productive developers... otherwise known as better games, and more games.

MIP Maps

This refers to a series of dumbed down versions of a texture that the graphics chip creates. They are used when a surface is so far away that it isn't big enough to show all the detail in the full texture. This saves processing power because smaller textures are easier to render, and it reduces aliasing artifacts often making games look better at the same time. Most next-gen systems and all current PC graphics boards support this feature. In some games you can clearly see where the MIPMAP drops to a lower level of quality (NFL2k). Trilinear filtering address this problem.

Bilinear filtering

All 3d game consoles use this except the original PlayStation. It creates the smooth look of textures as opposed to the blocky textures found in PlayStation games. It's sometimes known as texture anti-aliasing. The N64's bilinear filtering has become a bit dated at this time so you might want to look at newer consoles and PC graphics cards for a more accurate representation of the GameCube's filtering abilities. The GameCube tech demos would also do nicely (or the actual system now that it's out). Bilinear filtering works best with MIP Maps.

Trilinear Filtering

The next step after bilinear filtering. For the color of an individual pixel, not only does trilinear filtering blend together neighboring texture elements (texels), it also looks at the neighboring MIP Maps and blends those together. This requires two times the fill rate on most consoles, but GameCube does Trilinear filtering for free (although a few lazy developers have forgetting to implement it). Trilinear filtering requires the use of MIP Maps since it is defined as Bilinear with MIP Map blending.

Anisotropic Filtering

I'm not clear on exactly what the different is between trilinear and anistropic filtering, but I have seen it in action. One problem with MIP Maps is that they are always proportional to the original texture. This means that if the horizontal detail is very small, a low quality MIP Map will be used even though the verticle quality is very high. This will result in excessive blurring (just play a first person shooter and get as close to a wall as possible and look down the wall; if you're using bilinear filtering, you'll see the MIP Map jumps, with trilinear the wall will be smooth but quite blurry). Anisotropic filtering address this problem of texture blurring at extreme angles resulting in a much more refined image. Unfortunately, it's a fillrate hog and I don't think any GameCube games have used it yet (and they may never use it).

Real-time texture decompression (S3TC)

This feature is very important. It means that many textures will take up one sixth the amount of space that they would take up otherwise. This space reduction is maintained in every part of the system from the optical disk to the RAM. Other consoles might be able to compress textures, but they must be decompressed before use wasting precious RAM. The GameCube decompresses them as they are being used. Developers can take advantage of this by using more textures, better textures or a happy medium. Polygons are only half the battle when generating beautiful 3d worlds; the other half is textures. The GameCube will have plenty of room for textures, and as I will describe below above, it is going to do wonderful things with them. Unfortunately, not all textures can be compressed this way. Certain textures (anything with sharp changes in color like signs with words) look really bad with compression. These must still be handled normally.

Real-time decompression of the display list

This is related to geometry processing. In a normal console, the main processor spends about 90% of its time performing transforms leaving a meager 10% or less for things like artificial intelligence and physics. In the GameCube, the main processor "Gecko" will be completely free of this task as it is handled in the Graphics chip "Flipper". This was demonstrated dramatically in the Mario 128 tech demo. Miyamoto displayed 128 Mario's running, each with it's own complex AI. This wouldn't be possible if the "Gekko" was handling display lists. The Xbox's chip from Nvidia also has this feature. The PS2 doesn't have this very important feature.

HW motion compensation capability

Accelerated video rendering. If you have a DVD drive and a newer graphics card, then you have motion compensation. It's basically a fast way of decoding mpeg video and smoothing out the frame rates. This means that even though we can't watch DVD movies on the GameCube, our games can still feature FMV of DVD quality. The second party developer Silicon Knights plans to use quite a bit of FMV in it's up coming game, Too Human.

Multi-texture mapping

The GameCube can render eight textures in a single pass. This was probably the most stunning revelation at Space World. We certainly expected at least two textures, or even better, four textures per pass (like the Xbox); but eight!? This is absolutely great news. Some of the best-looking PC games to date are the direct result of multi-texturing. While multiple passes can achieve the same effects on the PS2, it requires much more processor time. So what does this really do for the look of a game, and what are some types of multi-texturing?

Currently, the most common use for a second texture layer is probably the light map. A light map combined with a normal texture gives the illusion of light shining on a surface. The Quake 2 and 3 engine use colored light maps extensively to achieve beautiful graphics, though the quality of those maps wasn't always very great.

Detail textures are another common use of multi-texturing. This is a high definition grayscale (or color) image that provides fine details on a surface. For example, things like wood grain, fine cracks in a concrete block, paint chipping off of a surface or the grout and chip marks on a floor tiles might be added by a detail texture.

Environment mapping is a technique used to produce reflections on a surface. It wouldn't be a mirror, but a shiny surface like glass or a new car. Even some dull objects, like lightly varnished wood, reflect their environment to some extent. You've seen Rare create working mirrors on the N64, so, normal reflections can be done as well, but that isn't called environment mapping.

Bump mapping is a little like detail texturing. A bump is actually a 3 dimensional thing, unlike wood grain. Proper bump mapping can create the illusion of fine 3d detail without additional polygons. Think of the bump map as a detail texture that constantly changes depending on the cameras vantage point. This creates the correct shadows around the "bumps" on the map giving a strong illusion of 3d. Check out Rogue Leader for the best example of bump mapping in a game. The third level Hoth (just skim the snow) is an especially good example. Now, notice that halfway through the level when you switch from your Speeder to an X-Wing you get a glimpse of the snow from straight up. Notice how different the bump map looks from this angle? That's the difference between bump mapping and detail texturing. The detail is dependant on the viewing angle.

As you can and will see, multi-texturing is very important to achieving beautiful graphics. The GameCube can do up to 8 textures per pass. Developers can choose to render any number of layers (from 0 to 8) per polygon to achieve effects in game. Rogue Leader is said to use at least two layers per polygon with many going to four or five and a few going all the way up to eight. Again, the PS2 can achieve multiple texture layers, but it will require a new pass for each layer (one extra polygon for each layer). PS2 gamers are going to see a big difference in the quality of GameCube games and PS2 games because of this. Only the best PS2 developers will be able to coax the PS2 into competing graphically with the GameCube, and the buzz around the industry is that a good GameCube game will still look better. XBox developers can also go up to eight texture layers, but it will require the rendering of an additional polygon.

Is that all of it?

After hearing about all these great effects, you might be surprised to learn that Nintendo apparently hasn't revealed all of the GameCube's abilities to us yet. There may be more graphical tricks up its sleeves. We can speculate and stare at tech demos trying to figure out what they are, but Nintendo has always been one to keep secrets well. Furthermore, GameCube's graphical engine is very customizable. As the years go by, we should see some very interesting use of GameCube's hardware (check out Star Fox Adventures for realistic fur for example). Is GameCube on the level with XBox? Only time will tell.

(FAQ Section by David Trammell)