A Buyer’s Guide to Graphics Cards

by 666isMONEY ☮ ♥ & ☠

There are literally hundreds of graphics cards on the market today, though they generally feature chipsets supplied by one of only two vendors: AMD/ATI and NVIDIA. AMD acquired ATI in 2006, and thus the names that are used when referencing chipset manufacturers are somewhat interchangeable at the time of this writing. Though chipsets manufacturers design a wide variety of chipsets, they do not make graphics cards for consumers. Instead, consumers buy graphics cards from any number of partners such as ASUS, MSI, or EVGA.

With both AMD/ATI and NVIDIA use many slightly different technologies; this is highly mitigated from the point of view of the average consumer thanks to industry standards such as OpenGL and DirectX. Standard such as these allow developers to quickly and easily produce software that will work on any supporting video card regardless of the actual architecture of that card. It is therefore important for anyone is looking to buy graphics cards to evaluate their software and only look for graphics cards that meet those minimum specifications.

It is also important to look at memory requirements. Graphics cards pair chipsets with high-speed memory. Memory comes in different sizes, speeds, bandwidths, and technologies. The easiest to understand of these is probably the size, which is usually rated in MegaBytes (MB) but is increasing at such a pace that it is not farfetched to guess that GigaByte (GB) will be the primary metric by or slightly after 2010. The more memory a graphics card has, the better it will perform when asked to render complex images and 3D environments.

While size is certainly important, so is speed. The faster the memory, the better the performance will typically be. It makes little sense for an average consumer to buy graphics cards with massive amounts of very slow memory. Memory speeds are usually rated as an ‘effective’ speed which is a combination of the actual speed, and the technology basis used. The technologies are part of the family of Dual Data Rate (DDR) memory, but the number that follows generally indicates how many times data is exchanged per clock cycle. For example: GDDR3 (Graphical DDR) is twice as effective at transferring data at the same time and at the same actual speed. Adding to this already murky world of memory speed is bandwidth, or how much memory can be transferred at one time. Some cards feature only 8 bytes (64 bits) while others use memory capable of transferring 64 bytes (512 bits) of memory at one time. The total speed of memory is a factor of actual speed times the memory bandwidth, times the technological basis.

Consumers looking to buy graphics cards also have to have a place to plug them into. For years the Advanced Graphics Port, or AGP for short, was the series of standards that were the most prominent. The current replacement for AGP is PCI-Express, which presently comes in special X16, or PCI-Express for Graphics (PCIEG) slots in both 1.0 and 2.0 specifications. 2.0 is twice as fast at taking information from the CPU as 1.0, and 3.0 is on the horizon.