 |
 |
 |
| titlegraphic |  |
The Videocard GuideLast updated 2005 Synopsis: The most important part to a videocard is its chipset. The amount of RAM and interface type aren't as important as the number of pixel pipelines and memory bus bandwidth. This guide also has information about various types of videocard features. One of my friends recently brought to light the confusion when buying a new videocard that a typical person might encounter. In response I created this guide to help remedy that situation. Marketing tends take advantage of consumers when it comes to technology, and so the following helps you become better informed about what to look for when it comes to videocards. What really counts? So oddly enough, the elements of a videocard that truly matter the most are generally the least often advertised! You often times hear about huge amounts of RAM on a card, or AGP 8x, or something of that matter. These things are really not as important as the marketers would like you to think they are. The true hierarchy is something like this: 1) Pixel Pipelines (Vertex Shaders) To explain it simply - this is the number of pathways (imagine big highways) that carry data from one part of the card to another. The more the better - though they sometimes come in different organizations. Vertex shaders are also pretty important. The best cards out there at the date of this writing have 16 pipelines - we'll see 24 pipeline cards come 2005. Really any reasonable card should have at least 4 pipelines - 2 pipelines is poor, 4 is moderate, 8 is good, 12 or 16 is excellent. 2) Memory Bus Bitwidth This is generally how cards are referred to as '64-bit' or something similar. 64-bit is terrible, 128-bit is reasonable, and 256-bit is good. This relates to the amount of data that can travel around the card. Like say a big highway with either 64 lanes or 128 lanes or 256 lanes. 3) GPU Clock Speed This is a bit complex to explain - but generally the higher the clock rate the better/faster the card is. You might compare it to processor speed in a computer in general. Different chipsets will have different ranges for speed. 4) RAM Clock Speed and Type The newer the RAM technology the more efficient and faster it is. SD-ram is bad, DDR ram is good, and GDR3 ram is excellent. Faster RAM is better, lower 'nano-second' ram (i.e. 4.5 ns) has better latency and is also better. 5) DirectX/Open GL Compatibility Generally cards with higher levels of DirectX and Open GL compatibility are a good idea. Get at least a DirectX 8.1 compliant card, and Open GL 1.3 or so. This is helpful both to making games look better, run higher quality games faster, and leaves headroom for upgrades. At the time of this writing the best out is DirectX 9.0c. Every two years or so they introduce a new DirectX version. 6) Amount of Memory This is not really as big of a deal as marketing wants you to think. The difference between a GeForce 4 ti4200 with 64 mb and 128 mb is very little, and only really impacts in games that really truly stress the card. The difference between a 128 mb Radeon 9800 Pro and a 256 mb Radeon 9800 Pro (with DDR2 too!) is also almost nothing, except in really extreme circumstances. Clearly the difference between a GeForce 6800 Ultra's 256 mb of RAM and the 64 on the ti4200 is huge - but that's also largely related to the fact that its is clocked at 1100 mhz, verses the 4200's 500 something mhz. Okay, point is, the amount of ram isn't that big of a deal. You'll see performance gains between identical cards with different amounts of RAM somewhere in the range of 2-5%. What really matters is adding the RAM in later generations. If a Radeon 9800 Pro with 64 mb of RAM existed, it would be significantly slower than the 128 version. What I'm saying is that videocards of a certain generation should have as much RAM as they can make use of, and no more. Guide by Generation 7) AGP or PCI Express Lastly is the interface. The speed difference between AGP 4x and 8x is about 5% in extreme conditions, and it's about 5-10% for PCI-Express cards for the time being. Basically better interfaces come with faster cards, which is why they might be faster. 8) Company The company (maker) of the card can also matter - different generations of cards from ATI and nVidia tend to be better than the other. Card makers like Asus manufacture quality cards that can be overclocked or will last longer. My personal opinion, which is debatable: Avoid early generation anything, especially TNT Rivas and Radeon 7200's. The GeForce4 MX generation is awful, don't touch "new" 64-bit cards with a 10ft pole. The 5x00 series by nVidia isn't very good when compared to the ATi cards in that range. Anything with an SE label is probably awful. Pricing... Lookout! ~75$ nVidia GeForce 4 ti4200 I wouldn't go higher or lower than those ranges. On-board
Video Woes What's DVI/S-video? Dual Monitors?
You get the idea - simply but you can do many more things at once, faster and easier. It's efficient, though costly. Anti-Ali-what?
Anisotrwa?
Now what's this other one? Anisotropic Filtering? Ever look at the floor in a game and see how it gets consistently blurry and blocky into the distance? Anisotropic filtering helps get rid of that. Here, another example:
So why do you want these things? Well they make your games/3D models look better. Do you need them to run games smoothly? No. Will they slow down performance when they're engaged? Yes. Do they help improve image quality and make a more life-like or cooler looking visual experience? Yes! Be careful when you turn these special features on - they can slow your card down considerably. In fact, with lower end cards it's not even really worthy to engage them at all. The ATi 9x00 series has superior AA and AF algorithms and methods when compared to the nVidia 5x00 run. The ATi x800/850's and nVidia 6800 series are almost the same. If you have more questions, go run games with various levels of AA and AF and see how well they work and how much better they look to you! Overclocking |
||||||||||||||
 |
||||||||||||||
 |
 |
