The acronym stands for "graphics processing unit," which is the integrated circuit responsible for handling 2D and 3D graphics, as well as video. It is the single most important component on any videocard. When it comes to PC games (and videogame consoles, too, for that matter), two manufacturers have pretty much divided in the market between themselves: ATI and nVidia.

Your motherboard's expansion bus will have one of two types of videocard slots: AGP or PCI Express. AGP (Accelerated Graphics Port) is the older and slower of the two standards (AGP 8x offers maximum bandwidth of 2.1GB/sec, compared to 8GB/sec for a 16-lane PCI Express), and it looks as though ATI and nVidia don't plan to support AGP with their newest parts. What's more, neither manufacturer's dual-GPU technologies support AGP. But if you're not ready to upgrade your motherboard, you will need to buy an AGP videocard.
If your motherboard does have PCI Express slots, you absolutely should buy a videocard with a PCI Express interface. PCI Express is a bi-directional serial protocol that's much faster than AGP. Each device on the PCI Express bus has it's own dedicated "highway," and each of these highways can be 1, 2, 4, 8, 16, or 32 lanes wide. Although current motherboards support only 16 PCI Express lans for graphic, the bandwidth available in such a configuration is roughly double that form an AGP slot.

GPUs (a.k.a. graphics processing units) create images by processing pixels in parallel pipelines. The more pixel pipelines a GPU has, the faster it will be capable of rendering an image. Everything else being equal, a GPU with 24 pixel pipelines will be faster and more powerful than one with 16 pipes; a 16-pipe GPU will be faster and more powerful than a 12-pipe GPU; and so on.

In a discussion of videocards, the word "core" refers to the videocard's GPU. Core clock speed measures how fast the GPU is set to operate and is stated in megathertz (MHz), or millions of cycles per second. Generally speaking, a faster core clock speed will translate to faster performance within a given GPU series. Comparing Radeon-series clock speeds to GeForce series clock speeds, on the other hand, is not necessarily a reliable measure of performance.

Videocards can be equipped with different types of memory in different quantities. The amount of memory on the videocard helps determine the maximum resolution the cards is capable of displaying. Budget videocards are generally equipped with 128MB of memory, but 256MB is the sweet spot. A few manufacturers are offering higher-end boards with 512MB, but at press time, those bad boys were still very bad news for our wallets

This specification measures how fast the videocard's memory is set to operate and is stated in mega-hertz (MHz) or millions of cycles per second. This spec can be confusing because some videocard manufacturers publish clock-speed specs while other promote "effective" clock speeds (even though they don't always label their spec that way). Allow us to explain: Nearly all videocards are outfitted with some flavor of DDR RAM (DDR is an acronym for "double data rate"). DDP memory is capable of transferring data on both the rising and falling edges of the clock signal; DDR RAM clocked at 500MHz, therefore, has an effective clock speed of 1GHz.

DDR2 RAM and DDR3 RAM operate on this same principle, but these types of memory are capable of higher clock speeds. GDDR3 RAM, meanwhile, is a double-data-rate memory designed specifically for videocard applications. It requires less power and generates less heat than DDR RAM.

The wider the connection between the videocard's GPU and its memory, the faster the GPU can process data. Budget videocards typically have a 128-bit interface to memory, while higher-end cards usually have a 256-bit interface.

Videocards have up to three types of connectors for connecting a display: DVI, VGA, or TV-Out. DVI (the acronym stands for Digital Visual Interface) is an Intel-developed standard for connecting a digital display, such as an LCD. VGA (the acronym ostensibly stands for Video Graphics Array, but the term is now widely used to describe any analog computer video signal or connectoe) connects an analog computer display. TV-Out, as its name implies, outputs an analog video signal compatible with consumer television, VCRs, and the like.

Most videocards have at least one DVI port and one analog VGA port. High-end cards offer dual DVI ports, so you can run two flat-panel displays in digital mode, but you can connect either connector using an adapter. Cards that support TV-Out usually provide a breakout cable with connectors for composite, S-Video, and, in some cases, component video.

The acronym stands for video-in/video-out, and it describes a port on the videocard that can both accept and produce a TV-compatible analog video signal. Videocards with VIVO ports can digitize analog video in real time, which makes them suitable for nonlinear video editing.

This is a collection of APIs (application program interfaces) developed by Microsoft to simplify fame development and eliminate hardware/software compatibility issues with the company's Windows operating system. As of this writing, DirectX 9.0c was the most current version. DirectX support is ubiquitous in games, but make sure the card you are considering supports the latest version.

This is a composite label for two components of Microsoft's DirectX technology. Pixel Shader Model 3.0 and Vertex Shader Model 3.0, which enable game developers to create more complex and realistic environments while simultaneously reducing CPU and memory overhead.

The limited numbers of games that currently utilize Shader Model 3.0 will revert to an earlier shader model when run on a GPU that doesn't support the new standard. NVidia's GeForce 6- and 7-series GPUs support Shader Model 3.0 now. ATI's R520 supports the technology too.

The picture on the left shows the effect by Shader Model 3.0