![Old](images/statusicon/post_old.gif)
01-16-2007, 07:22 PM
|
![mike_15's Avatar](DSC05462_u_2830_dateline_1288066645.JPG?u=2830&dateline=1288066645) |
Sage Advanced User
|
|
Join Date: Nov 2004
Location: Pittsburgh
Posts: 158
|
|
Video Card Info
I am shopping for a new video card again and found myself a bit less knowlegable on the video card data than I would like so I did some searching and found some info I though was worth sharing.
How to pick the perfect graphics card
Graphics technology, SLI and Crossfire explained, plus see full results for over 80 cards
Rory Reid, Personal Computer World 15 Feb 2006
http://www.pcmag.co.uk/actions/trackback/2150316
http://www.pcmag.co.uk/personal-comp...-graphics-card
Quote:
AGP is perfectly capable of handling games with complex graphics as it has a maximum throughput of 2.1Gbytes/sec.
...
As a result, graphics card manufacturers have begun developing cards that use the PCI Express bus, which offer speeds of up to 4Gbytes/sec.
|
Quote:
GPUs and memory
The speed of the GPU in your graphics card is also important. This varies from card to card, with low-end models typically clocked at around 300MHz, and overclocked high-end cards reaching up to 800MHz.
Like CPUs, the number of transistors on a GPU helps define the processor’s level of performance.
The speed at which these transistors (miniature electronic switches) turn on and off is dictated by the GPU’s clock speed.
The more transistors a GPU consists of, and the faster they switch on and off, the faster your GPU is likely to run.
|
Quote:
While most integrated graphics cards will borrow from main system memory, mid-range and high-end graphics cards use a type known as Graphics Double Data Rate (GDDR) memory.
This is more expensive than standard DDR Ram, as it has lower power and heat dispersal requirements that allows the memory chips to run faster.
Most graphics cards quote their memory as having an effective clock speed, which is usually twice the memory’s actual clock speed.
This effective clock speed is the result of double data rate (DDR) memory sending two chunks of data over the memory bus during a single clock cycle (double pumped).
Whereas standard DDR (sometimes called DDR1) memory designed for motherboards runs at an effective maximum of 400MHz, faster DDR2 memory runs at up to 533MHz, and the fastest type of GDDR memory, currently GDDR3, runs at around 750MHz with an effective speed of 1,500MHz.
Having high-speed memory isn’t the only concern. As the size and complexity of images in games increases, so does the need to have a fast and reliable means of transferring those images between memory and the GPU – particularly when running games at high resolution.
The amount of memory bandwidth available to a graphics card is calculated by multiplying the size of the memory bus by the speed of the memory.
So a card using a memory bus that is 256bits (32bytes) wide with an effective memory clock speed of 1,500MHz would provide a memory bandwidth of 48Gbytes/sec.
It’s all fairly complicated, but what you need to remember is that, all else being equal, the bigger the bandwidth, the less chance there is of having a bottleneck in the graphics card, and the faster your high-resolution games will run.
|
Quote:
Pipelines and shaders
When a new graphics cards is released, many reviewers will probably relate the card’s number of pixel shader pipelines to its relative performance. So what are they and how do they work?
A pixel shader is a graphics function that calculates effects on a per-pixel basis. They allow programmers to control the lighting, shading and colour attributes of each pixel in any scene.
Whereas old-style graphics cards could only apply effects to whole entities such as single polygons or textures, pixel shaders can bring a greater amount of surface detail to graphics and provide great freedom to games developers to help realise their artistic vision.
The more pixel shader pipelines a graphics card has, the faster it can apply complicated visual effects to a scene.
Vertex shaders are similar to pixel shaders, except that they work on a vertex-only basis, applying effects to the corners (vertices) of polygons in a scene. They are particularly useful in complicated animations involving facial expressions or skeletal movement.
|
Quote:
In our experience, games using OpenGL run slightly better on Nvidia hardware, whereas DirectX games tend to fare better on ATI cards.
|
|