Basics of The Graphics card
The Graphics card is responsible for delivering the image you see on your PC monitor. Its GPU (Graphics Processing Unit) processes the data and changes it into a signal to the monitor. There are many factors to a graphics card. Choosing one can be a tricky business these days as there is so much technology that is different in each new graphics card release. If your looking for a graphics card that is to be used for hi resolution gaming or computer aided design (CAD) or even Photoshop use then picking a card can be quite tricky as many of these types of graphics cards are expensive and you need to get it right first time. In this article we will look at the different aspects of the
graphics card, some of the technologies involved and what they do in order for you to make a good decision on your purchase. This is a beginners guide, for more information some of the sections contain links to more detailed information.
When PC’s first came and for some time after, the graphics cards purpose was only to display the image on the screen. The amount of memory you got on a graphics card was very small and was not needed to a great extent. Today’s graphics cards do more than just display an image, they help the processor with the job of processing when it comes to the graphics. The graphics card would in effect accelerate the process of displaying the image on screen.
This was needed when the 3D gaming world took the center stage. The speed required to process the images on screen at 60 frames per second and process the code for the game itself was simply too much for a CPU to handle on its own and o the games would simply crawl along at a very slow pace. The graphics card would use some of its own built in instruction logic to added things such as textures and lighting effects, fog effect and bump mapping to give a far more detailed picture. Also the speeds of graphics cards have improved a great deal in order to let these effects be used without the problem of the frame rate dropping.
One of the biggest improvements to graphics technology was nothing to do with an increase of speed or efficiency but with graphical improvements. Anti-aliasing was technology to allow jagged edges of computer sprites to be smoothed on screen by blending colours. Monitors have pixels which are rectangular by nature and are incapable of drawing a diagonal line without looking crooked. Anti-aliasing technology doesn’t prevent this from happening but uses tricks of the eye in order to make images look smoother and more pleasing to the eye. Lots more information on Anti-Aliasing can be found in our Anti-Aliasing article.
Anisotropic filtering is a method of increasing graphics card performance by allowing the graphics card to render textures in the background or further away at a lower quality level. There are many
levels of Anisotropic filtering (AF) usually 1x, 2x, 4x, 8x, and maxed at 16x. The higher the multiplier the better the textures will look in the background but will increase the performance hit on the graphics card. For a more detailed look at Anisotropic filtering you can see this article by Will Harris – Anisotropic filtering
The refresh rate of a graphics card is no different to that of a monitor, it is the amount of times per
second the image is “refreshed” and is measured in Hz (60Hz =60 refreshes per second) With a graphics card however the refresh rate is the amount of times a full image is calculated ready for display. If for example your set your graphics card to 100Hz the it would attempt to calculate a new image 100 times per second. Fine on the surface as you you would think the faster the better, but remember you also have a monitor that needs to display this image. If your monitor is only capable of displaying 75hz then you will have frames rendered before the monitor was ready to display them, this causes screen tears or unwanted “artifacts” on the screen. To avoid this, you should enable the V-Sync feature – short for Vertical Synchronisation, this feature limits the graphics card to the refresh rate of the monitor even if it can render the image faster.
AGP or PCI Express
Two types of Graphics card available today are the AGP and PCI-e versions. The AGP (Accelerated Graphics Port) is the older of the two technologies but still quite popular as many people still have these slots incorporated into there motherboards. The PCI-Express (Peripheral Component Interconnect) version has been around for a few years now and new graphics cards and motherboards alike are using this technology. PCI express offers a greater scope for data transfer to and from the graphics card and main memory. If buying a graphics card today then the PCI-Express is the way to go as AGP cards are dying out. PCI Express graphics card use a much more powerful bus system which offers more bandwidth capability and has the extra benefit of being able to prioritize data to the devices that need it the most, and a lot of the time the graphics card will have that priority.
How do you measure the speed of a graphics card?
Measuring the speed of the graphics card is a lot more difficult than with the CPU or RAM or even the hard disk. There are many factors which affect how quickly the graphics card can do its job. Many of these only come into play when the graphics card is undertaking certain tasks. While each of the details below are important, none of them alone can determine the real speed of your graphics card, this is because overall performance is determined by all of the factors, not only this
but different graphicscards will be more efficient at different tasks that others. However comparing the numbers between graphics cards will give a decent indication of the speed and performance compared to other cards in the range.
Core clock speed – Much the same as the way you measure the speed of a CPU.
The core speed of the Graphics card is measured in MHz or GHz and represents the
amount of clock cycles the graphics process can do per second. This is a good but not definitive way of telling how fast the graphics card is. When comparing graphics cards in terms of clock speed its best to compare cards
that are in the same range, that way you can be sure that the faster clock speed will be the faster card.
Memory clock speed – Exactly the same of as the core clock speed, except of course that it is for the memory of the graphics card and not the core. This is just as important as the core speed as the memory contains textures that need to be applied to the pixels. Memory Clock speed is important because it affects the Memory Bandwidth. A higher memory clock speed results in greater ability to apply Anti-Aliasing and Anisotropic Filtering. Memory Clock speed is often measured in DDR rate (Double Data Rate), So a memory clock speed of 500MHz DDR has an effective speed of 1GHz.
Pixel Pipelines – The amount of pixel pipelines a graphics card has can have a great impact on the speed of the image rendering. This is all about pixel pushing power. A card with 8 pipelines can process twice as many pixels as a card of the same core speed and 4 pipelines. Again this is not
the definitive calculation for the amount of pixels a card can push each cycle but will give you a good indication. More pixels that can be displayed equals higher resolution that your graphics card can use for gaming etc.
Textures per pipeline – This only come into effect when multiple textures are needed on the one pixel. Simply put if a multiple texture is needed, then a graphics card with more textures per pipeline will be quicker. On single textured pixels the amount of textures per pipeline will have no effect.
There are more smaller things such as T&L technology anti-aliasing and various other quality increasing and speed increasing technology that different cards have. I wont go into them all here as there are a great many between all the cards on the market, but all of these technologies help in one way or another and can mean that a graphics card with slightly lower clock or memory speeds for example can still be the faster card.
Amount of Cores/Shaders/Streamprocessors – This area of speed measurement can only really be used when comparing graphics cards from the same range. The reason being that different GPU’s have different types of Cores/shaders/streamprocessors. I write all 3 together because they are essentially the same thing with different names. They are all GPU cores in effect but because some core are more focussed on specific tasks than others they were given different titles. This is the same reason why we cant simply measure core count across graphics card ranges. One GPU may well have less cores and so less overall processing power but the cores could be more flexible in the tasks that they were designed for. This would make the graphics card more efficient and actually more powerful when it comes to certain tasks. having more cores seems like the better option and it is in some cases but it is not a black and white situation.
The memory bandwidth is the rate of data transferred from the GPU to the graphics memory. This has been one of the biggest bottle necks in graphics cards for years. Newer cards are overcoming this problem with expensive memory solutions at high speeds. The higher the memory bandwidth the better the graphics card will be able to retrieve data and textures from the graphics memory. As this is a real bottle neck this is a really important feature of the graphics card.
When it comes to measuring the speed of a graphics card however you cant do better than studying benchmarks. All the spec sheets in the world cant tell you how the graphics card performs in real world tasks. So check out the benchmarks on various hardware sites to see how they perform in different games, in graphics design or whatever you need your graphics card to do. You may be surprised to see that the cards with the most raw power don’t always come out on top of all the benchmarks.