@
star89 what @
kreacher has said is the essence of the difference;
videos and other pre-rendered clips are basically codecs which contain the relevant data in the form of a basic algorithm that the CPU OR a GPU simply needs to decode and read.
So let me give you a more general elaborate flow chart.
WHAT happens when you start to play a video file?
You locate the video you want to play --> double-click on it [this initiates the appropriate program / protocols] --> the CPU engages with the program in question --> the program in question helps the CPU / GPU decode and reintegrate the data blocks that are formed onto the screen as images / films / video clips.
In this process the CPU / GPU is just reading information and applying it to the situation, RAM usage is not high because the RAM is just a transit point for data and hard-drive usage is moderate because DATA is being read of the same.
What happens when you play a game?
You initiate the game, initially a few cut-scenes and logos are displayed, these are examples of pre-rendered media, not very unlike the videos discussed earlier.
Once you enter a game, the CPU needs to read and in a few instances even write data on the hard-drive [save games, checkpoints]. Now unlike the videos which are data chunks encoded into a certain format, a game is an application unto itself --> so the CPU start loading the map OR immediate map items / textures / lights et al. Now this puts up a wall of calculation, something that was not happening while viewing a simple HD video --> after loading all this data onto a canvas [and doing the major PHYSICS algorithms], the CPU forwards this to the graphics card which starts applying its own calculations and stores ready frames on its buffer V-RAM --> now as you have a dynamic camera in a game, the CPU needs to take that into account and keeps doing minor correction for light, draw distance and it stores all this in RAM until the graphics card deems it necessary to be accessed [RAM usage heads north] --> as you interact with the word and vice-versa the CPU also has to continually update your character statistics [more work for the CPU] --> as you go from different lighting scenarios you also make it harder for the graphics card, infact simulating textures like water [not looping pre-rendered textures] take a lot out of the graphics card and depending on detail can even bring the most powerful cards onto their knees.
Few factors that further compound problems of playing games --
- There were earlier two distinct types of shaders, Vertex-shaders and Pixel-shaders and depending on the graphics card these were a fixed number. Since the advent of Dx 10 in the market this was revised in the favour of a common pool of Geometry-shaders / Stream-shaders / -processors. That depending on the scene will decide how many of these will act as Vertex-shaders and the rest as Pixel-shaders. This has made the entire scenario more balanced and more efficient.
- The dedicated V-RAM of a graphics card will offer a boost in performance if your scene is very large and thus large textures must be stored relatively close on hand to be accessed. This does not mean an entry-level graphics card with an inflated V-RAM cache will perform a mid-range card with smaller RAM cache. This is because the latter will address its memory management in a more efficient manner.