Video Cards

Without 2D video circuitry we wouldn't be able to see anything on our monitors. Video circuitry is either integrated directly into a system's motherboard, or located on a video card that plugs into the motherboard. Integrated or "on board" video is cost effective, but that's because manufacturers nearly always integrate low-end video. This guide will focus on graphics cards (also called video cards or video adapters).

The type of card you need depends on what you do with your computer. If you primarily use your computer for business and web surfing, a good, "dedicated 2D" card can produce exceptional looking documents and presentations. If you're a gamer or someone who frequently works with graphics, you will probably want a card that can handle serious 3D.

Understanding the Types
There are a few different kinds of video cards. The simplest is the frame buffer type card, which was king in the days of text-based applications, but is no longer widely used. This type of card relied on the CPU to do all of the graphics calculations. As graphics and multimedia began to play a bigger role in computing, the workload became too much for the CPU; that lead to accelerator cards. Accelerator cards handle the graphics calculation "grunt work," while your system's CPU directs the show and assists with certain tasks.

While accelerator cards are still widely used, the most recent and expensive type of video card has what is called a co-processor or GPU (graphics processing unit). A GPU acts similarly to a CPU except that it is dedicated to graphics work. The GPU does the majority of the video related calculations, freeing up the CPU to attend to other matters. A GPU is actually more complex than a CPU, and hence more expensive.

It is also important to know that there are dedicated 2D cards as well as 2D/3D combination cards. All modern 3D cards are also 2D capable, and are therefore commonly referred to as 2D/3D combination cards. Combination cards may be of either the accelerator variety or the co-processor variety. Dedicated 2D cards will be of the accelerator variety, and should speed up performance in a Window's environment. For instance, lines, windows, curves, and fonts will be drawn faster. Contrary to what many people seem to think, if you only have dedicated 2D, you will still be able to view 3D graphics. That's because everything is 2D based. However, 3D graphics won't look nearly as good, plus performance may lag.

In addition, both AMD and Intel incorporate media enhancing technology into their processors, such as 3Dnow! (AMD), SIMD2 (Intel), and MMX (both). These technologies won't allow you to view graphics if you don't already have graphics circuitry, but if you do, and the software you are using supports the enhancement technology, images should look a bit better than they otherwise would.

Resolution
Resolution refers to the number of pixels (smallest unit of a video image) that can be displayed on a screen at once. A pixel is not a fixed size. The more pixels there are, the higher the resolution, and the sharper the image.

A monitor can usually support a number of different fixed resolutions, such as 800x600 dpi (dots per inch), 1024x768 dpi, or 1280x1024 dpi, where the first number represents the number of pixels horizontally, and the second, the number of pixels vertically. To a certain degree, the size of your monitor determines which resolution you should use.

When shopping for a video card, make sure it can support the resolution you want to use at the appropriate color depth (we'll discuss color next). Here are some basic resolution guidelines.

Professional 2D graphics users will probably want to use a resolution of 1600x1200 dpi or higher, which usually requires at least a 19" monitor.
Professional 3D work requires a slightly lower resolution--1280x1024dpi is usually enough. You'll probably want at least a 17" monitor, but a 15" will support that resolution.
2D business graphic users with a 17" or 19" monitor and 32-bit (true) color should opt for 1024x768 dpi or 1280x1024 dpi resolution.

Color
The more bits used per pixel (bit depth), the better the image's color detail will be. For a display to fool the eye into seeing full or true color, 24 and sometimes 32 bits per pixel are required.

*Note: High color results in only a very small loss of visible image quality, but may give you slightly faster performance.

RAMDAC
RAMDAC stands for "random access memory digital-to-analog converter" (you may also see this term referred to simply as DAC.) Your computer processes information digitally, but your monitor only understands information in analog format, (unless you have a LCD flat panel monitor, which understands digital information). RAMDAC translates digital information into analog information, then sends it through the video cable to the monitor.

The RAMDAC affects both image quality and refresh rate (which we cover in the next section). You can judge a RAMDAC's quality by its speed. Today speeds range from approximately 200MHz to 350MHz. Faster is better.


Refresh Rate
Refresh Rate refers to the number of times per second that the image on your display is recast. If the refresh rate is too low, you will notice screen flicker. For most people 72-80 Hertz is great. However, if you sit at your computer for long periods of time, you may want to consider a card with a higher refresh rate as that will help reduce the strain on your eyes.

Tip
The higher the resolution, and the more colors you use, the more memory and processing power required. If your computer is having trouble keeping up with the demands of a graphic intense application, try lowering your resolution and switching to high color verses true color. This should make a difference unless it is your CPU, motherboard, or memory slowing things down, and not your 3D card. The ability of a powerful 3D card will be wasted in a computer with a weak CPU and insufficient memory.

Another way to improve performance is by lowering your refresh rate.

Interface
Your video card's interface is what connects it to the computer's main memory. The main memory is where the CPU fetches the data and instructions it needs to perform tasks.

The latest interface is called AGP (Accelerated Graphics Port). Your motherboard may have an AGP1x/2x slot, an AGP4x slot, or no AGP slot at all. If you don't have an AGP slot, then you'll have to purchase a PCI video card. AGP has a much higher bandwidth than PCI (the higher the bandwidth, the faster data can be transmitted), which makes it preferable. However, a PCI card is just fine for most business applications--just not for 3D games etc.

AGP requires software support from your operating system, graphics card, and graphics drivers. Windows '98 has built-in support for AGP, but Windows '95 does not.

Video Memory
A video card's memory has a big impact on its performance. First, it is necessary to make sure there is enough memory for what you want to do. Here are some basic guidelines.

For professional graphics, look for a card with anywhere from 32MB to 128MB RAM (or higher), depending on your needs.
For 2D images, the amount of graphics RAM isn't as important as a high-speed RAMDAC to handle high refresh rates with high resolutions. But, it's probably best to get at least 16 MBs since cards with less usually have older chipsets.
For 3D graphics, on the other hand, the more RAM the graphics card has, the better. You won't want to go lower than 32MB.
It is also important to make sure the memory on your card has a high enough bandwidth. The memory's bandwidth is the rate at which data can flow through it, and is measured in MB/sec. Basically, as graphic performance increases, so does the amount of video memory and bandwidth you need.

Before an image shows up on your screen, all the data that goes into creating it must first pass through the video memory. The more color, texture, and effects there are making the image look realistic, and the higher the resolution, the more data there is that must pass through the video memory. A larger bandwidth allows more data to pass through the video memory per second. Memory Bandwidth = speed of the memory * width.

Over the years there have been numerous types of memory used on video cards. Today, however, nearly all PC video cards have either SDRAM or SGRAM memory. The width of the memory will be either 32, 64, 128, 256 bits, or even higher. It's not uncommon to find the memory's bit width listed in the video card's title.

In addition, you may notice that many high-end 3D cards have DDR (Double Data Rate) SDRAM or SGRAM. If a video card has DDR it means the card transfers data on both the rising and falling edge of the clock cycle, and therefore has double memory bandwidth. Also, Direct Rambus DRAM (RDRAM), a new all-purpose memory developed by Rambus Inc., may also start showing up on graphics cards.

Application Program Interfaces (APIs)
Graphics technology is probably the fastest growing area of the PC industry. New chipsets come to market about every six months. This presents a problem for software developers, because there's no way they can customize their software for each chip.

This dilemma lead to the use of Application Program Interfaces or APIs. The software vendor writes the program's graphics commands in standardized code for the API--rather than for the chip itself. Then, the graphics-chip-maker writes driver software, which translates the software's standardized code into a format that is easier for the graphics chip to understand.

In the mainstream PC world, two 3D graphics APIs are dominant--OpenGL and Microsoft's Direct3D. If you plan on playing 3D games, it's a good idea to make sure your card supports both APIs--that way, you should be able to play almost any PC game.

Be aware, that there may be a version of a particular game written specifically for your video chip, rather than for an API. This version will give you better performance than an OpenGL or Direct3D version. But the trade-off is that you won't be able to play that game with any other graphics chip (for instance, on a friend's computer, or if you upgrade to a video card with a different chip).

The Importance of Drivers
As just mentioned, once the program has imparted its graphics code to the API, the graphics chip's integrated driver software will translate that code into the chip's "dialect," so to speak.

The driver software is written by the chip-maker. Some chip-makers write very efficient drivers, and some don't. Well-designed drivers can improve the performance of a graphics card to the point that a mediocre card with a great driver can outperform a high-end card with a poor driver.

It's a good idea be certain you are using the most recent driver. Chip-makers are constantly updating drivers to ensure that their chips will function properly with new applications. In addition, driver updates may fix bugs. Often, updating your chip's driver is as easy as going to the manufacturer's website and downloading the latest version. But if you run into trouble, try visiting driverguide.com.

3D Graphics
A video card teams up with a system's motherboard/chipset, CPU, memory, and PCI or AGP interface, to put 3D graphics onto the computer screen. The first stage of the process used to be called Geometry, but now it's commonly referred to as Transform and Lighting or T&L. The geometry used to be done completely by the CPU, but now some chipsets such as ATI's Radeon, and Nvidia's GeForce2 and GeForce3 families have transform and lighting engines enabling them to perform much of that work themselves. A video card with a T&L engine, or similar technology, will be a much better 3D performer than a card without.

Stage 1: Transform and Lighting
When a program instructs the computer to put a 3D image on the screen, the computer starts with the first frame and calculates where the vertices and lines should be. These configurations of vertices and lines are called polygons (or triangles). Currently, a medium range 3D card has a throughput of 20 million triangles/sec. The higher the triangle count, the smoother the images will appear on your screen.

The next step is lighting. Shading, reflection, shadows, and other lighting effects are added to the scene depending on an object's position relative to the light source.

In addition, when a character moves in a 3D game, the T&L engine is also responsible for recalculating the position of all the objects in the image.

Stage 2: Rendering
The graphics card is in charge of the rendering stage. During the rendering stage, the "wire-frame" image created during T&L is filled in with color and texture.

A number of different techniques may be used during the rendering stage. The more techniques used, the more realistic the image appears. But the down side is that it's more work for your card and CPU, and could affect performance. When the system can't keep up it may cause the frame rate to decrease. The frame rate is the number of frames that can be displayed in a certain amount of time. An excellent 3D card can display 60 frames/sec. even at high resolutions (6/01).

If the frame rate gets too low, then movement will appear choppy--a problem referred to as dropped frames. Some video card makers deliberately pick and choose which 3D techniques to include, in an effort to steer clear of dropped frames.

Here is a list of 3D techniques you are likely to see mentioned.

Gourad Shading: Helps define the shape of 3D objects, giving them the appearance of depth.
Clipping: Determines which parts of objects will not be visible on the display so that the processor doesn't waste time with things that the user won't even see.
Transparency: Makes things look transparent.
Texture Mapping: Creates textures such as wood, plaster, or stone.
Dithering: Tricks the eye into seeing more colors than are actually present. Fewer colors mean fewer computations.
Fogging: Fogging blurs objects that are in the distance so they look more realistic. In addition, blurred objects are quicker to compute because they are less detailed.
Filtering: Filters, such as the "bi-linear" filter are used to clean up and smooth out images that otherwise might look blocky.
Anti-aliasing: A technique used to reduce the noise (speckling) present in an image.
Bump Mapping: Makes surfaces look bumpy.
Z-buffering: Calculates the pixels that will be loaded into the frame buffer.
Alpha Blending: Blends two bitmaps together making one appear transparent.
Interpolation: Smoothes out the image, making it look realistic even up close.
MIP Mapping: Stores several versions of a textured bitmap to prepare for instances when the user moves close to an object. Without MIP Mapping an object will blink and sparkle as the processor continually recalculates the image. With MIP Mapping the texture of the object will get larger making the user feel as though they are getting closer.
Perspective Correct: This makes a textured 3D image look realistic from any angle.
Texture Mapping: This puts a bitmap onto a 3D surface to make the object look more realistic.

Special Features

The following are a few popular special features you might want to keep in mind when you're shopping for a video card.

DVD Acceleration: If you have a DVD drive on your computer, or plan to get one, this feature will improve the quality of playback. The information on a DVD disk is highly compressed--that's part of the reason why they can hold so much. Video cards with DVD acceleration have special hardware that improves their ability to process highly compressed information, which means smoother playback with few or no compression artifacts.

TV Out: TV Out refers to a video card's ability to display its signal on a TV or VCR. If you have a DVD drive but don't like watching movies on your computer, this feature will allow you to transfer the picture to your TV. It's also great for gamers, and handy for giving presentations.

Multiple Monitors: This feature enables you to connect a second monitor and work on different tasks at the same time while keeping both in view.


Manufacturers & Chips
Your video card's chipset is its most important part. It's the brain. Knowing the type of chipset on the card you are considering is important to making a good decision. The name of a graphics card's chip is almost always in the card's title. However if it's not, when you check the specifications remember that manufacturers call chips by several different terms including chipset, engine, accelerator, co-processor, graphics processor, and graphics processing unit (GPU).

Currently, some major names in the graphics chip business are Nvidia, Matrox, and ATI. Nvidia calls their chips GPUs (graphic processing units). They sell their GPUs to card manufacturers such as Gainward, MSI, Asus, Visiontech, Daytona, and others. However, ATI and Matrox produce cards under their own names. 3Dfx, maker of the popular VooDoo line of cards, was bought out by Nvidia in 12/00, however their cards are still on the market.

The following are the primary graphics chip families, listed by manufacturer. In some cases we've listed the manufacturer's description of the chip as found on their website, to help give you an idea of what's available.

ATI
Radeon Series: ATI's most powerful chip yet. The Radeon chip offers ATI's best 3D performance. This series includes:
Radeon 8500 Technology
All-In-Wonder Radeon 8500DV: All-In-Wonder cards offer features such as TV-on-demand, which lets you watch and control TV on your PC.
Radeon 8500: "With 64MB DDR memory and intelligent architecture, RADEON 8500 gives you the new level of character realism and texture versatility, raising the bar in visual quality and performance." -- ATI Website
Radeon 7500 Technology
Radeon 7500: "RADEON 7500 is a powerful and versatile graphic solution. 64MB of powerful DDR memory along with the RADEON 7500 GPU provides high performance acceleration of today's demanding 3D graphic applications. Industry leading DVD playback, support for dual independent displays, and support for digital flat panel (DVI-I) monitors meet the needs of a wide range of home and business graphic users." -- ATI Website
Radeon Technology
Radeon VE Dual Display Edition

All-In-Wonder Radeon

Radeon 64MB DDR version
Radeon 64MB SDRAM version
Radeon 32MB DDR version

Radeon 32MB SDRAM version
Rage Series

Rage 128 Pro Technology: "Designed for mainstream 2D/3D graphics and multimedia acceleration."-ATI website. This family includes:
All-In-Wonder 128 Pro
Rage Fury Pro
Xpert 2000 Pro
Rage 128 Technology: "A graphics and multimedia accelerator that offers leading-edge performance in all three vectors of visual computing: 3D, 2D, and video."-ATI website. This family includes:
Xpert 2000
All-In-Wonder 128
Xpert 128
Rage Pro Technology: This family includes:
Xpert 98

Matrox
Millennium G550 Series: "The new Millennium G550 lets you communicate in 3D over the Internet, while providing outstanding image quality and flexible display support, including renowned Matrox DualHead and superior DVD playback." -- Matrox Website
Millennium G450 Series: This chip doesn't offer the most extensive array of 3D features, so hard core gamers may want to opt for something different. However, it does have very high-end 2D performance, as well as Matrox's proprietary Dual Head technology (supports two displays), making it ideal for business use. This series includes:
Millennium G450 w/ 16 MB DDR SDRAM
Millennium G450 w/ 32 MB DDR SDRAM
Marvel G450 eTV: Matrox's "total entertainment solution."
G400 chip
G200 chip
Nvidia
GeForce4 Family: "NVIDIA introduces the groundbreaking, top-to-bottom GeForce4 family of GPUs—delivering new levels of graphics performance and display flexibility to desktop and mobile PCs." --Nvidia website
GeForce4 Ti Series
The ideal solutions for: Advanced 3D applications, Extreme gaming, HDTV/DVD playback.
NVIDIA introduces the GeForce4 Ti–delivering the most radically immersive graphic environments ever seen on a desktop PC. Groundbreaking features including the advanced nfiniteFX™ II Engine, superior Accuview Antialiasing™ techniques and flexible nView™ multi-display technology, deliver screaming performance and fierce processing power–for truly unmatched ferocious graphics power. --Nvidia Website

GeForce4 Ti 4600
Vertices per Second: 136 Million
Fill Rate: 4.8 Billion AA Samples/Sec.
Operations per Second: 1.23 Trillion
Memory Bandwidth: 10.4GB/Sec.
Maximum Memory: 128MB

GeForce4 Ti 4400
Vertices per Second: 125 Million
Fill Rate: 4.4 Billion AA Samples/Sec.
Operations per Second: 1.12 Trillion
Memory Bandwidth: 8.8GB/Sec.
Maximum Memory: 128MB
GeForce4 Ti 4200
Vertices per Second: 113 Million
Fill Rate: 4 Billion AA Samples/Sec.
Operations per Second: 1.03 Trillion
Memory Bandwidth: up to 8GB/Sec.
Maximum Memory: 128MB
GeForce4 MX Series
The ideal solutions for: Multiple display & desktop flexibility, 2D office applications, Web surfing.
GeForce4 MX 460
Fill Rate: 1.2 Billion Texels/Sec.
Triangles per Second: 38 Million
Memory Bandwidth: 8.8GB/Sec.
Maximum Memory: 64MB

GeForce4 MX 440
Fill Rate: 1.1 Billion Texels/Sec.
Triangles per Second: 34 Million
Memory Bandwidth: 6.4GB/Sec.
Maximum Memory: 64MB

GeForce4 MX 420
Fill Rate: 1 Billion Texels/Sec.
Triangles per Second: 31 Million
Memory Bandwidth: 2.7GB/Sec.
Maximum Memory: 64MB
GeForce4 Go Series
The ideal solutions for: Mobile business presentations, Gaming on the go, Multiple display flexibility.
GeForce4 440 Go
Memory Bandwidth: 7GB/sec.
Fill Rate: 880M texels/sec.
GPU Core Clock: 220 MHz
Package Size: 31x31mm
GeForce4 420 Go
Memory Bandwidth: 3.2GB/sec.
Fill Rate: 800M texels/sec.
GPU Core Clock: 200 MHz
Package Size: 31x31mm
Quadro Family: This family of cards is intended for the creative professional. These cards are designed for workstations rather than PCs.
Quadro DCC
Quadro2 Pro
Quadro2 MXR/EX
Quadro
Quadro2 Go
*Personal Cinema: "Transform your PC into a complete entertainment system" -- Nvidia Website. This new offering from Nvidia is not a graphics card on its own, but is desigend to be used in conjunction with any GeForce2 or higher GPU. Personal Cinema brings Intelligent TV, and video capture/editing/publishing to the feature sets of the aforementioned cards.
GeForce3 Family: "..shakes up the gaming industry with unprecedented visual effects and sizzling frame ratesù-injecting life into the previously artificial world of computer-generated graphics." -- Nvidia website (regarding GeForce3)
"The newest members of the GeForce3 family GeForce3 Ti 500 and GeForce3 Ti 200 bring two new features to end users: shadow buffer technology and 3D textures. These new features, combined with the nfiniteFX Engine and the Lightspeed Memory Architecture, deliver the most visually compelling and complete graphics experience available today. No other technology provides as much functionality. That's why the GeForce3 Ti series is the reference platform of choice for both the Microsoft DirectX 8.1 and the SGI OpenGL application programming interfaces (APIs), as well as the technology foundation for the Microsoft Xbox game console." -- Nvidia website
GeForce3 Ti 500
Graphics Core: 256-bit
Memory Interface: 128-bit DDR
Fill Rate: 3.84 Billion AA Samples/Sec.
Operations per Second: 960 Billion
Memory Bandwidth: 8.0GB/Sec.
GeForce3
Graphics Core: 256-bit
Memory Interface: 128-bit DDR
Fill Rate: 3.2 Billion AA Samples/Sec.
Operations per Second: 800 Billion
Memory Bandwidth: 7.36GB/Sec.

GeForce3 Ti 200
Graphics Core: 256-bit
Memory Interface: 128-bit DDR
Fill Rate: 2.8 Billion AA Samples/Sec.
Operations per Second: 700 Billion
Memory Bandwidth: 6.4GB/Sec.

GeForce2 Family:
GeForce2 Ti: "GeForce2 Ti, the latest member of the GeForce2 family, redefines graphics for mainstream computers, delivering 1 billion pixels/sec. rendering power and producing an amazing 6.4GB/sec. of bandwidth. The GeForce2 Ti provides consumers with unprecedented visual experiences at mainstream price points. " -- Nvidia website
GeForce2 Ultra: Although the GeForce2 Ti is the latest addition to the GeForce2 family, the GeForce2 Ultra offers the same memory interface (128-bit DDR), the same triangles/sec (31 mill/sec), and the same pixels/sec (1 bill.), but also has a higher memory bandwidth (7.36GB/sec vs. 6.4GB/sec).
GeForce2 Pro: "GeForce2 Pro shares the same features as the GeForce2 GTS but offers 20% more bandwidth."-Nvidia website
GeForce2 GTS: "The original per-pixel shading GPU, GeForce2 GTS provides a high-performance graphics, gaming, and HDTV (high definition TV) solution."-Nvidia website
GeForce2 MX: "The mainstream GPU-ideal for work or for play. In addition to providing incredible performance, GeForce2 MX is the only product in this family that provides TwinView multi-monitor support and Digital Vibrance Control."-Nvidia website
GeForce2 MX 200
GeForce2 MX
GeForce2 MX 400
GeForce2 Go: "The world's first mobile GPU. GeForce2 Go provides notebook users with the power to create, present, and entertain-anytime, anyplace."-Nvidia website
GeForce 256: The original graphics processing unit.
TNT2: "...was the first chipset to offer a 32-bit frame buffer for better quality visuals at higher resolutions, 32-bit color for more realistic colors, and a 32-bit Z/stencil buffer for incredible 3D effects-blazing the trail for graphics processing units (GPUs) such as GeForce 256 and GeForce2 GTS." -- Nvidia website
Vanta: "...a low-cost, 128-bit TwiN Texel, highly integrated 3D graphics processor, designed specifically for value-conscious users. Leveraging the NVIDIA TNT architecture, NVIDIA Vanta delivers breakthrough 3D and industry-leading 2D graphics performance, making it ideal for today's basic desktop PC. The NVIDIA Vanta comes in 4-16MB configurations." -- Nvidia website

3Dfx
Voodoo5
Voodoo4
Voodoo3

Conclusion
Remember that the video card industry moves very quickly, so it's not a bad idea to check the latest industry news at sites like Hardware Central (hardwarecentral.com), Tom's Hardware (tomshardware.com), or chip makers sites: ati.com, nvidia.com, and matrox.com. In addition, once you have a few video cards in mind, read some reviews. A great place to look is neoseeker.com, which archives reviews from sites across the Web. But remember that a video card won't necessarily behave the same way on your system as it did on the reviewer's. Much depends on your motherboard, CPU, and memory, as well as the application you are running.

If you need any extra help, consider visiting an online community forum; again, Hardware Central, and Tom's Hardware have forums where folks always seem willing to offer their opinions and advice. In addition, there are step by step installation guides available at sites across the Web, such as the "Video Card Installation Guide" at Sharky Extreme (sharkeyextreme.com).

Congratulations! You're ready to shop for a new video card!