CAPSULE INFORMATION

On each graphics part, I'll list the following information; this section is a summary of what each of those pieces of information means.

Shader Model:

There are three different shader models that we'll see in modern notebook graphics parts: Shader Model 2, 2b, and 3. The higher the shader model, the increased number of graphic features that part supports. This is sort of esoteric, but it means the difference between "High Detail" and "Very High Detail" in games like Far Cry and Age of Empires III. A Shader Model 3 part may allow you to add additional details to parts of a game that supports it, provided the part itself has enough horsepower to keep the game running playably. (To wit: nVidia's GeForce FX series all supported Shader Model 2, but the FX Go 5200's performance was too poor to actually run games using it.) Note that Model 2b is a strictly ATI standard, and is very minimally supported.

Pipelines:

The number of pixel pipelines (shaders) a given GPU has. What these do is generally esoterica; they essentially refer to the number of pixels that can be shaded per cycle in the chip. More is generally better, but the bare minimum is typically 4, though 8 pipeline parts and greater are becoming a lot more common. Again, this line is being blurred and this section may very well be obsolete by next year's article.

Memory Bus Width:

This is a specification that can have an incredibly profound effect on performance. A bigger bus width allows more data from the RAM to be transferred at once. Bus width comes in 64-bit, 128-bit, and 256-bit, and generally 128-bit is the minimum you want. What's different now as opposed to last year is that nVidia's low end TurboCache parts have 64-bit bus width, but still offer very acceptable performance.

Specification Clock Speeds:

When a graphics part is released, it typically has a specification for how fast the part and its memory run. On integrated parts, the memory clock won't be listed as these parts (again, excluding certain Radeon Xpress parts) share system memory. While I didn't list these last year, I've noticed companies have tightened up and standardized these clocks a bit more, so I'm a little more comfortable producing these. These are REFERENCE, however. The MacBook Pro's Mobility Radeon X1600 is notoriously underclocked, while the Mobility Radeon X600 in my own notebook runs 10MHz over specification on the memory. There is still some variance.