about vid cards

[skateis2s]

hello, can someone please explain exactly the two clock speeds of video cards, and what is AF? and what the difference between 4x and 8x, isnt it bandwidth and what is the actual difference for having 128megs of vid memory rather than 64megs of vid memory, how does that actually help? thanks! i just want to understand vid cards better !

 

[MournSanity]

Originally posted by: skateis2s
hello, can someone please explain exactly the two clock speeds of video cards, and what is AF? and what the difference between 4x and 8x, isnt it bandwidth and what is the actual difference for having 128megs of vid memory rather than 64megs of vid memory, how does that actually help? thanks! i just want to understand vid cards better !

One clock speed is for the core of the graphics card, the graphics processor basically. The other clock speed is for memory. AF stands for Anisotropic Filtering, which I think sharpens the textures in game or something.

 

[skateis2s]

right on man, thanks!

 

[modedepe]

1. The clock speeds are of the gpu and of the memory on the card.
2. AF = anisotropic filtering, which basically sharpens the images and makes them less blurry.
3. The difference between agp4x and 8x is the amount of bandwidth (~2.1Gb/s vs ~1Gb/s). Although this may seem like a large difference, it translates into very little real world performance increase. The amount of bandwidth which agp4x provides is hardly ever even used up. Furthermore, when the video card must use the bus to transfer data at all a large fps penalty occurs, simply because there is a large amount of latency.
4. Having more memory on the card keeps the card from having to transfer data across the agp bus, which as stated before slows things down tremendously. For example, if your card has 64mb of memory but you need 90mb for the textures in a game, you're going to have to use the agp bus to transfer the textures from system memory. Higher amounts of memory are necessary also when you're running higher res or aa/af.

 

[skateis2s]

perfect thanks!!!

 

[Ionizer86]

And in current scenarios, we find that 64MB GF4 cards do as well as 128MB GF4 cards, and 128MB Radeon 9800 cards do as well as 256MB Radeon 9800's. I don't think many games can slow down a 64MB card unless you run with massive resolution and massive AA/AF.

 

[titananandtech]

The two speeds listed are Memory speed and VPU/GPU speed. Just like a faster CPU runs programs faster a faster VPU/GPU will be able to draw more pixels per second. If the VPU/GPU is too fast then it will spend all its time waiting for textures to be read from memory, which is why you want fast memory too. In addition to memory speed the memory can be hooked up to a 64bit, 128bit, or 256bit bus. A 100Mhz 256bit bus would have the same bandwidth as a 400Mhz 64bit bus. Wider is better 🙂

AF is anisotropic filtering. It makes textures look better on angles. Check out these links.
http://developer.nvidia.com/object/Anisotropic_Filtering_OpenGL.html
http://developer.nvidia.com/object/Filter_Anisotropic.html

4x and 8x AGP are how much faster the AGP port is compared to a standard port which they haven't made since 1997 or so 😉 A lot of video card companies are using the AGP memory for vertex information and using the video memory for textures. This means all your verticies are streamed over the AGP bus. Normally this isn't a problem for most games as not many push a lot of polygons because they have to work on cards without hardware T&L units (Radeon IGP, Intel Extreme Graphics, PowerVR Kyro), but in games with high polygon counts like Homeworld2 this can become a bottle neck. 4X AGP is more than good enough for now and even the near future, but AGP 8x is more future proof.

Or if you're referring to 4x vs 8x anisotropy filtering that is how many times it reads the texture. The more times it reads the texture the more accurate it can filter it so it is less blurry. Of course the more it reads the texture the slower it draws...

The more memory your video card has the more textures it can hold. If you have a scene with 50 megs of textures in it then there is no difference between a 64meg card and a 256meg card at all. If you have 70megs of textures then each frame the video with only a little bit of memory is going to have to copy some of those textures over the AGP bus (and then having AGP 8x is even more helpful!). If you have 150megs of textures each frame each texture will have to be copied over the AGP bus. Currently there are a ton of 32meg video cards that must still be supported by game companies so 128megs is going to be enough for a long time.

So basically an otherwise identical card supporting AGP8x instead of AGP4x and with 256megs instead of 64megs will perform exactly the same in most games on the market today. But as more advanced games come out it won't be as future proof.