The Gamer's Graphics & Display Settings Guide

[Page 4] The Graphics Process (Pt.3)



Step 7 - The GPU Processes the Data


The GPU now accesses the data in VRAM and undertakes a range of complex graphics calculations in what is commonly referred to as the Graphics Pipeline. This 'pipeline' has several stages, the aim of which is to repeatedly convert raw 3D data into something viewable on a 2 dimensional display screen:


1. Transformation - We start off with raw Vertex data, a vertex being a point in 3D space which, when several are joined together later will make up Wireframe 3D shapes called polygons. In the first stage of the pipeline, this raw 3D vertex data is transformed in such a way as to create and place all 3D objects correctly into what will be a 2D space - your monitor. Everything is scaled and positioned correctly relative to each other. A Vertex Shader may be used at this point to apply certain effects such as warping of objects.


2. Lighting - All the objects are then lit correctly based on the location of the light source(s) and any reflective surfaces. The data is still in vertex form though, so it's basically still just lots of points with different shades of color based on the lighting applied.


3. Viewpoint - The objects are then further adjusted to account for the fact that they will be viewed from a particular angle and distance. So objects closer to the virtual viewer are made bigger, objects further away smaller, and everything is rotated and moved around appropriately.


4. Clipping - Objects which would not be visible to the viewer based on the current Field of View (FOV) are essentially removed. For example, if you have a FOV of 90 degrees, which is typical for most games, the remaining 270 degrees of objects around you which you can't see at any one time will be removed. This is done to reduce the amount of wasteful calculations the GPU would otherwise have to do for something that's not visible.


5. Triangle Setup - At this point all the vertices (points) are joined together to create triangles or polygons, i.e. multi-sided wireframe shapes, which is what we are more familiar with when picturing basic 3D graphics on a computer screen.


6. Rasterization - A Pixel is the smallest unit of graphics information displayed on your monitor. Since everything in a 3D scene will end up being displayed on a flat 2D surface as a pixel, every object in the scene is assigned a particular pixel location on the screen.


7. Culling - This process picks up where clipping left off; every pixel which would still not be visible is removed: the rear side of objects that are facing you, objects obscured by other objects, etc. are all removed as they are effectively not visible to the viewer at the current time and hence not worth wasting additional computational power over.


8. Texture & Shading - At this point, each pixel making up the image has any relevant colors or effects applied to it. A texture is a 2D image which is added to a 3D object's surface to make it look realistic. A Pixel Shader may combine a range of color, depth and texture effects to relevant pixels to create special effects like wrinkled, glossy or transparent surfaces on objects.


Performance Tip: A modern graphics card actually has multiple pipelines operating at the same time and this 'multithreading' allows the graphics card to undertake the various stages listed above in parallel and thus perform them much more quickly. As expected, graphics cards with more pipelines generally perform better. It is only relatively recently that graphics cards have been able to render detailed 3D graphics in real-time fast enough to display smoothly. In the past, complex 3D scenes (e.g. for movie special effects) were rendered slowly over a period of sometimes hours or days, whereas now they are done dozens of times a second. Gaming is one of the most calculation-intensive things you can do on a home PC, so do not underestimate how much stress an entire PC is put under when playing a complex 3D game. It's not just your graphics card either which gets a workout. The more recent the game, the more complex it will usually be, and hence the greater the likelihood that it brings out latent instabilities anywhere on your system and causes crashes and lockups.



Step 8 - The Data is Sent to the Frame Buffer


The data from the steps above - also known as a single Frame - is sent for temporary storage to an area of VRAM called the Frame Buffer. As the name suggests, this area buffers (stores) frames, ready to send them through to your computer monitor as required.


There are usually two buffered images held at any one time, a process called Double Buffering. One completed frame is held in the Primary (or 'front') buffer, while another one is being composed in the Secondary (or 'back') buffer. The Primary buffer is the one whose output is displayed on your screen at any time. While the contents of the primary buffer are being displayed on the monitor, the graphics card is preparing another frame in the secondary buffer, and as soon as your monitor is preparing to refresh itself again, the buffers are 'flipped' so that the newly completed content of the secondary buffer is then displayed while what was the primary buffer (now secondary) empties itself and composes another new frame. This Page Flipping technique helps prevent a potential bottleneck and increases performance when displaying frames rapidly in a row in 3D applications.


The larger the resolution and the more complex the image, the more data it has and hence the more space it will take up in the frame buffer. If any additional effects are being applied to the frame, like Antialiasing to smooth jagged lines, or Anisotropic Filtering to clear up distant textures then that will in turn again increase the size of the image (frame).


Performance Tip: Usually only two frames are held in the frame buffer. However Triple Buffering is a technique which provides a third, or Tertiary buffer in the frame buffer, improving performance when Vertical Synchronization (VSync) is enabled, because it gives the graphics card more room to work on completing frames while it is waiting for the monitor to signal that it is ready to receive a new frame. However this also understandably uses up more VRAM and thus can increase stuttering and hitching. Furthermore, it can also introduce control lag and graphical glitches - for more details see the Triple Buffering section.



Step 9 - The Image is Sent to the Screen


The prepared image is now ready to be displayed, sent from the Frame Buffer through a RAMDAC (Random Access Memory Digital to Analog Convertor), up the video cable to your monitor. The RAMDAC is necessary to convert digital information (i.e. information in 0's and 1's) in your VRAM to an analog format (i.e. video signal voltages) displayable by most monitors. A RAMDAC constantly converts frames to viewable images many times a second.


The RAMDAC is important because it affects both the maximum possible refresh rate at particular resolutions (in conjunction with the monitor's capabilities), and the image quality on the screen. Fortunately most recent graphics cards have good quality fast RAMDACs, so this is not a major concern - the monitor's refresh rate is usually the limitation, not the RAMDAC's.


Importantly, while a traditional CRT monitor is dependent on a RAMDAC, an LCD monitor is not reliant on them if you use a pure digital connection such as DVI (Digital Video Interface) between your graphics card and monitor. Using the DVI input on an LCD monitor results in the RAMDAC being bypassed, since the original digital signal travels unaltered to be displayed on your LCD screen, whereas using a standard VGA connection on an LCD requires that the digital signal be converted to analog (by the RAMDAC) and then converted back again to digital by the LCD monitor before being displayed, which is not as efficient. Note that using a DVI to VGA converter for a CRT monitor also effectively bypasses the RAMDAC, as the output is digital, but the digital to analog conversion still occurs in the cable itself.


Performance Tip: As this step highlights, LCD monitors provide the best image quality when connected using DVI, as there is no likelihood for data degradation through repeated conversion. However there may be reasons to use an analog connection on an LCD monitor - see the Response Time section for more details. If you have a CRT monitor and an old graphics card, and believe your image quality is not the best it can be, the RAMDAC on the old card may not be of sufficiently good quality. This is particularly noticeable for old graphics cards which produce fuzzy 2D image quality, such as a noticeably blurry Windows Desktop, and less distinct 3D quality at higher resolutions. This is definitely not good for your eyes.



That covers the basics of how 3D graphics work on your system. Hopefully you now have a much clearer understanding of how your system works when playing a 3D game, and hence you should be better armed at both troubleshooting any problems and working out the best ways to improve performance. Remember though that a large range of important performance tips are contained in my relevant game and system tweak guides, such as the TweakGuides Tweaking Companion, so please make sure to go through those as well as reading the advice above.



The next section delves into the details of a range of common graphics settings which gamers often encounter. Understanding precisely how these work is once again critical to both troubleshooting and performance optimization.