When shopping for a video card, it's important to know what to look for so you can get the best card for you money and not end up with a lemon or something that doesn't work in your system. Heading out on the web or to a store to shop for a video card can be a daunting experience...there's a lot of different things thrown at you and not all of them really matter. Marketing departments the world over have spent a lot of time making purchases like this as confusing as possible, so you'll spend more money on less. I'll try to unravel things a little bit.
Parts of a video card in order of importance -
1. GPU -
the Graphics Processing Unit, or brain of the video card, is what actually does all the number crunching and creates the pixels that end up on your screen. It's the integrated circuit chip that sits near the center of the video card and is almost always covered by a large heatsink and fan. Without any doubt, this is the MOST important thing on the video card. The GPU architecture's generation, sub-generation (where it falls among its siblings), and its speed are all important factors in choosing the right card.
2. Memory speed -
I'm separating things when it comes to memory, because the size of the onboard memory on a video card is one of the most common ways that the marketing departments trick you into buying something you don't need. The speed of the onboard memory, and the width of the bus (memory bandwidth) are extremely important... the size of the memory is of MUCH less importance.
3. Power requirements -
Newer video cards require a lot of stable electrical juice in order to pump out those frames. It's very important to know what power supply you've got (or what kind you are getting) and that it is capable of providing enough clean and stable power for whatever video card you choose.
4. Cooling -
Video cards are not all created equally, although a lot of them are very similar. Look carefully at the cooling solutions on the card. Hardware review sites will very often discuss cooling as part of their reviews, and even do objective testing... you can also use customer reviews on a site like Newegg as an additional data point. If everyone is saying "This thing is LOUD", it probably is.
Some cards are much louder than others, some are much more effective (often the loud ones
). Some are just flat out annoying and don't work well. Newer video cards put out a tremendous amount of heat...and a good heatsink and fan on the card is essential... too much noise is annoying, but too little cooling means the card and the whole rest of the computer turn into a space heater.
5. Memory size -
Now here we are... memory size can be significant in high-end cards. If you are getting the newest-gen GPU, and you've got a ton of horsepower under your graphics card's hood... a large memory buffer may be useful. 512MB or even more RAM can come in handy when you are pushing the settings all the way to max in games, and using a high resolution with effects like anti-aliasing turned on. In other words, larger memory helps in the high end.
In the mid-range and low-end, more memory is often a very deceptive way of selling you a slower video card. Manufacturers know that you will look at memory size as a primary criteria for picking out a card, so they will take a mid-range or low-end GPU and slap a whole bunch of very slow RAM on it. Sadly, the GPU doesn't have the horsepower to utilize that much RAM (you can't turn the settings high enough and get acceptable frame rates), and even if you could, the RAM is slow and will hold back performance.
So if you are looking at two cards with the same GPU, but one has 256MB of RAM and the other has 512MB... look at the SPEED of the memory. It's the easiest way to spot a crappy deal in that situation.
In number 1A above, I mentioned GPU generations and sub-generations. The naming schemes behind modern GPUs are extremely confusing, and it changes often enough that no explanation is going to be completely accurate for very long. The only way to keep up is to constantly read tech reviews and articles, or else ask for help. What follows is a general description of the way GPUs of different generations and sub-generations are usually named, to get you started.
Terms defined -
A GPU "generation"
refers to a processor architecture (the internal design of the graphics chip). Some generations involve only minor changes and additions to the GPU architecture, others are very significant, including complete redesigns and re-layouts. The most common graphics programming interface (The software foundation that games are built on) in use today is from Microsoft, called Direct X. This is not just a set of program tools for creating 3D content, it also lays out guidelines for how hardware should be designed in order to take advantage (and properly accelerate) the software. Direct X also handles sound and other things, but we're just going to talk about graphics here. OpenGL is the other popular API.
You will often hear GPU generations referred to based on what DirectX version they support, or are compliant with. This is significant because at least for future games which utilize DirectX, the features that your card supports may dictate the games that your computer can play. If your video hardware supports the features of DirectX 9.0c (is DX9c compliant), it will not be able to utilize the features of DirectX 10. However, with very few exceptions, DirectX hardware is always backwards compatible. So...DirectX 10 compliant hardware will definitely still work great with DirectX 9, 8, etc.
A GPU "sub-generation"
refers to where a particular graphics chip falls (in terms of performance and features) compared to it's siblings of the same generation. Obviously, with each new graphic processor design, there are multiple versions for different price points. These are differentiated using a sub-generational naming scheme.
A generational "Refresh"
is usually an update at the mid-point between generations. This usually involves minor updates and improvements to the architecture, possibly more memory, and higher clock speeds.
. Graphics cards are also named using a suffix (usually some letters) at the end to denote where they fall within a sub-generation. Again, this is generally based on relative performance...but there are a lot of games played with suffixes, so be careful. Just like generational refresh, the naming suffix may include information about the type, size, and speed of memory...or other features of the entire graphics card, not just the GPU itself.
For AMD/ATI, the current naming scheme looks like the following -
An example -
The first X means nothing, in terms of the GPU or card...it's a marketing tool (X is a cool letter that makes you want to buy things). The GPU generation in this case is 1xxx, or 1000. The sub-generation here is 9xx, or 900. For AMD/ATI, this denotes the highest-end GPU in this generation (there are also 800, 700, 600, 500? and 300 sub-generations within the 1000 generation, with decreasing power and price respectively). The suffix in this case is "XTX", which denotes the highest-end card within the sub-generation 900. Other suffixes AMD/ATI commonly use are (in order of highest to lowest) - XT, GT, Pro. Suffixes will change and shuffle around a LOT, so take this info with a grain of salt.
So, what we have with the X1900XTX is the highest-end part in the generation (highest sub-gen and highest suffix), i.e., a very powerful card.
When AMD/ATI does refreshes, they have (at least with the last few) broken the number down even further. So, a refresh part would carry a "50" number... example - X1950XTX.
For Nvidia -
An example -
The generation is the first number, in this case 8xxx or 8000. This is the latest generation from Nvidia. The second number is the sub-generation, 8xx or 800 (other current sub-generations are 600, 400, and 200). 800 denotes the highest-end part in the generation. The suffix "GTX" denotes the highest-end part within the sub-generation... Nvidia will typically use these suffixes (in general order of performance) - GTX, GTS, GT, pro/XT. Nvidia will also sometimes, depending on the generation, have cards with no suffix. Nvidia has been using the same naming scheme for quite a long time (with minor variations), and in another couple generations they may have to change to something new.
Nvidia will often use the sub-generation number to also denote generational refreshes. They typically start a new generation using the even-numbered sub-generations (200, 400, 600, 800) and then use the odd-numbered generations to denote the refresh (300, 500, 700, 900). This can get very murky...but you can expect that an 8900GTX would be the refresh version of the example card above.
Check my links page
for great hardware sites that provide more information and reviews/comparisons of graphics cards. Also, if you aren't sure, don't hesitate to ask!
Now, it's time for a quiz. In your head (don't post the answer), figure out where this card falls in ATI's lineup -
Have fun shopping, and let us know what you're getting.