Time to update a sticky

Shadow ÒÓ said:

AGP Accelerated graphics port. The standard PC interface for videocards. The current spec is AGP 8x, which provides about 2.1GB/s of bandwidth between the videocard and the motherboard.

AGP Aperture Size The amount of system memory an AGP videocard can use for 3D information.

AGP Fast Writes Enabling AGP Fast Writes lets a program access the videocard directly, instead of through system memory. This reduces bandwidth demands on the bus.

AGP Pro A beefy version of the AGP interface designed to deliver extra power to workstation-class videocards. AGP Pro uses a slightly longer AGP slot, and comes in two flavors, 50-watt and 110-watt. Current consumer accelerators from ATI and nVidia use an onboard Molex connector to draw additional power, because the AGP Pro interface isn't ubiquitous.

Alpha Channel Every pixel in a game is described by four color channels: R, G, B and alpha. The first three channels store data on the pixel's mixture of red, green and blue, while the alpha channel describes a pixel's relative opacity. In 32-bit integer color, the alpha value is contained in the last 8-bits of the pixel's color info.

Anisotropic Filtering An advanced filtering technique that corrects textures's perspectives within this texture is drawn on an object that stretches deep into a sceene.

Antialiasing Any technique used to smooth jagged edges that are created when straight lines are drawn in a 3D game or program.

API Application programming interface; basically a set of rules and routines that developers can use to design games and applications. APIs provide programmers with a set of pre-defined function calls that can be used for basic tasks, such as drawing windows and buttons, or advanced operations, like applying a bump map to a polygon.

Bilinear Filtering The most basic filtering strategy, bilinear filtering reduces texture pixelization by blending adjoining textures at their seams. Modern 3D accelerators use bilinear filtering by default.

CG nVidia's high-level shader language. It includes all the functions of Microsoft's high-level shader language (HLSL), as well as some functions specific to nVidia GPUs.

Core Clock The speed at which a 3D card's graphics chip (GPU) runs, usually measured in megahertz (MHz).

DB-15 The standard 15-pin connector for analog monitors. If you use a CRT or analog LCD monitor, this is the connector you use.

DDR Double data rate, usually used to describe memory. DDR memory can transfer two chunks of data every clock cycle, a "double pumped" strategy that effectively doubles memory speed.

DDR-II The latest version of DDR memory. DDR-II memory doesn't transfer more data per clock cycle than original DDR memory, but it has been modified to run at much higher clock speeds than just plain DDR.

Direct3D Microsoft's 3D API. Includes all the function calls a developer needs to program old-style fixed-function games as well as new programmable shader games.

DVI The connector used for digital LCD monitors: effectively eliminates the video noise from digital-to-digital conversion.

DVI-D A DVI hardware spec that only includes support for a digital signal. DVI-I cables will not fit a DVI-D connector.

DVI-I A DVI hardware spec that allows both analog and digital signals to be passed out the DVI port.

Fixed-Function 3D chips running fixed function 3D pipelines can execute a limited number of functions very rapidly. Unfortunately, fixed function pipelines are also inflexible, and developers can't easily use any functions that aren't built into the chip. Programmable shaders pipelines are replacing fixed-function pipelines.

Floating-Point Color Floating-point numbers use decimals (e.g. 5.78) to describe a value. When games and applications use floating-point numbers to describe color, rendering errors are minimized, and colors become richer and more true to life. Videocards with DirectX 9 support must be able to use high-precision, 128-bit floating-point numbers to describe colors.

GPU Graphics processing unit; the main chip on a videocard. The GPU performs all the calculations that draw 3D images on your screen. GPUs include a rudimentary level of programmable shader support, whereas VPU's are totally programmable.

Hardware Transform and Lighting A graphics chip's hardware "T&L" engin can perform lighting calculations very quickly. Prior to the introduction of hardware T&L, these calculations were performed by the already overloaded CPU.

High-Level Shader Language HLSL, the shader language that is included with DirectX 9. High-level shader languages allowe game developers to use a C-like programming language to write shader programs for games and graphics applications. Without an HLSL, developers would have to use a confusing, hardware specific assembly level language.

Integer Color Integer numbers have been used to describe color information since 3D graphics first appeared on teh PC. Compared with floating-point numbers, integer numbers provide a very limited palette of colors. Integers are whole numbers (e.g. 4).

Memory Bandwidth The amount of data that can be transferred between a graphics chip and its onboard memory. Measured in megabyttes per scond (MB/s). To calculate memory bandwidth, multiply your memory bus bit-width by the speed of the memory and the number fo chunks of data transferred per clock, then devide that number by 8. For example: For 128-bit DDR memory running at 500 MHz, you would multiply 128x500x2 then devide the product by 8. The result is 16,000MB/s or 16GB/s.

Memory Bus Width The size of each data chunk that the memory bus can transfer each clock cycle. Usually measured in bits.

MIP-Mapping mip-mapping provieds several different levels of detail for every texture in a scene, and places a low-quality, high-res textures in the foreground (where game players focus their attention). Because low-res textures consume relatively little memory, mip-mapping can help speed up frame rates. Unfortunately, without filtering, demarcation lines can occur where different mip-map levels intersect.

Multi-Sampling An antialiasing technique that draws only the extra pixels needed to antialias jagged lines. Each GPU manufacturer uses a different algorithm to perform multi-sampling, so it's difficult to do an apples-to-apples comparison with anti-aliasing enabled. multi-sampling is the preferred AA technique.

NV30 The code-name for nVidia's first DirectX 9 GPU, which is the core for the GeForce FX line. It operates at speeds up to 500 MHz, and is fabbed on a 0.13 micron process. The 5800 and 5800 Ultra use NV30 core, while the 5200 series uses the NV34 core, and the 5600 series uses the NV31 core.

OpenGL A 3D API designed by an independent standards body. OpenGL differs from DirectX mainly in it's extensibility; it can be changed and added to between major revisions to take advantage of new hardware features.

PCI Express The next generation PCI replacement, slated to appear in late 2003 or early 2004. PCI express will also replace the AGP bus with PCI Express for Graphics, which should deliver 4GB/s of bandwidth to an from the 3D accelerator.

Pixel Shader A programmable shader unit that manipulates individual pixels. Pixel shaders can create bump maps, reflective surfaces, and even programmatically generated textures.

Process In respect to microprocessors, the term refers to the size of transistors on a chip. Formerly measured in microns (one millionth of a meter), but now measured in nanometers (one billionth of a meter). A smaller process technology lets chips run at higher speeds and lower tempratures.

Programmable Shader Programmable shader units are the areas of GPUs that can run generalized programs rather than fixed 3Dfunctions. there are two basic types of programmable shader units: pixel and vertex.

Programmatically Generated Texture Traditional textures in 3D games are created in a bitmap editor, such as Photoshop, whereas programmatically generated textures are described in real-time using shader programs. the newfangled textures do not require trilinear or anisotropic filtering in order to maintain the proper perspective when they're applied to a 3D model.

R300 The code-name for ATI's first DirectX 9 core. Fabbed on a 0.15-micron process, runs up to 325MHz, and supports a 256-bit memory pipeline. The R300 is used in the Radeon 9500, 9500 Pro, 9700 (and Pro and All-in-Wonder Pro).

R350 ATI's follow-up to the R300. Fabbed on a 0.15 micron process, but operates at speeds up to 380MHz. The 9800 (and Pro and All-in-Wonder Pro) use the R350 core.

R360 The updated for for the 9800 Pro and XT models.

RAMDAC Random access memory digital to analog converter. The RAMDAC converts the digital signal that is output from your 3D card's frame buffer into an analog signal that your CRD monitor can understand. Your videocard's maximum resolution and refresh rates are determined by RAMDAC strength.

Renderman A high-level language for the complex shaders that are used in very detailed animated movies (such as Shrek). Renderman shaders use multiple computers working in tandem to draw a single frame. It's a very long and laborious process, but the 3D hardware companies hope to run Renderman-like shaders on sonsumer gaming hardware (at 60 fps) within the next few years. Full Renderman shasers can run on DirectX 9 cards, but not at full speed.

SDRAM Synchronous dynamic random access memory. SDRAM is the dominant form of memory on today's videocards. Its low cost and high speed make for a perfect combo. Variants of SDRAM (e.g. DDR SDRAM) run as high as 500MHz and provide more than 20GB/s of memory bandwidth.

Super-Sample An antialiasing technique. Super-Sampling renders a scene at high resolution then filters it down to the desired size. While this produces a terrific AA effect, it's very slow, especially compared with multi-sampling. Super-Sampling isn't used on modern 3D accelerators.

Texture Compression Most modern APIs include some form of texture compression. Because most textures are simply bitmaps, they're easily compressed. This conserves the limited bandwidth between system memory and video memory, and between video memory and the GPU, with virtually no degradation in image quality.

Trilinear Filtering A texture filtering technique that blends pixels in different mip-maps to hide the seams between different mip-map levels.

Vertex Shaders A programmable shader unit that specilizes in geometry calculations. Vertex shaders handle terrain in morphing, some lighting calculations, and some shadow calculations.

Vertical Sync A setting that allows you to synchronize the frame rate of your 3D accelerator's output to the refresh rate of your monitor. Disabling vsync may allow your frame rate to go higher than your monitors rate, but at the expense of nasty horizontal tears across the full screen image.

VPU Visual processing unit. Some 3D companies designate their 3D chips as "VPUs". VPUs are fully programmable. In fact, in theory you could perform any calculation on a VPU that would run on a CPU. On the other hand, GPUs are capable of just a limited number of sequential operations.

Volumetric Texture 3D textrues that can describe the rings of objects like tree rings. Volumetric textures consume significaltly more space (and thus more memory bandwidth) than normal two-dimensional textures, but can be effeciently generated using pixel shader programs.

Z-buffer The Z-buffer controls an objects distance from the camera in a 3D rendered scene. It's also called the depth buffer. Low precision Z-buffers can make objects seem to pop, or clip through each other.

Click to expand...