The Triple 32

What is the TNT2 Ultra?

With the recently released TNT2 and other chipsets, it's no wonder that people's heads are spinning with 3D terms and reference specifications for all the different chipsets. Let us first define nVidia's reference TNT2 Ultra chipset, and what it has to offer.

32 x 3

The answer is not 96. The 32 x 3 defines some of the impressive reference specifications for the TNT2 Ultra chipset. These three "32s" are RAM, 3D rendering color depth, and the Z/stencil buffer. Among these features, the "vanilla" TNT2 chipset differs only in that it comes with 16 MB of SDRAM instead of 32 MB. Now, let us explain these features in a little more depth.

32 MB SDRAM

The onboard video memory, or frame buffer, is where a full screen's worth of data is stored, or buffered. Then, this information is converted by the RAMDAC into a form useable by the monitor, where it is displayed. As 32-bit 3D rendering becomes a playable option in newer 3D games, extra memory must be devoted to rendering, in order to prevent your framerates from slowing down to a crawl.

You see, in the frame buffer, one screen of information is stored at a time. This screenful of information is stored as an array of pixels. Each pixel has some bits assigned to it, depending on what that pixel is supposed to looks like. In 16-bit rendering, 16 bits are assigned to each pixel to "describe" the color of the pixel. Moving up to 32-bit rendering adds 16 more bits to the pixels, in order to display a larger number of colors to increase realism. As an example, moving from 1024x768x16-bit in 3Dmark99 MAX to 1024x768x32-bit increases the required amount of texture memory by ~5 MB. So in a very complex 3D scene, 32-bit rendering can incur quite a need for memory. Having a larger frame buffer allows more bits to be assigned per pixel, and also allows more pixels to be rendered at one time, which is the resolution. Thus, we get higher resolutions for games, rendered in 32-bit color.

32-bit rendering

32-bit rendering describes the color depth that is used for each pixel. With 16-bit rendering, you get a total of ~65,000 colors. With 32-bit rendering, the number goes up to over some 16 million, with bits left over for alpha values (which determines transparency) and such! Having increased color depth allows more realistic looking images, because colors blend into each other better, allowing for smooth gradients. Otherwise, areas where colors blend together will look "spotty", as you see dithering.

24-bit Z-buffer + 8-bit stencil buffer

The Z-buffer is part of the video memory that is reserved for another purpose. The Z-buffer determines which objects on the screen are supposed to appear behind other ones, so it stores depth information. When the Z-buffer determines that an object is behind another one, which is just determined by a value given to the pixels, it will not be rendered. That way, objects in the front are displayed properly, and rendering time is decreased.

The stencil buffer is a buffer that is devoted to special effects. The stencil buffer can create such effects as plain or volumetric shadows, silhouettes, and dissolves. By devoting this buffer to these effects, the card's engine is relieved of rendering these things.