The performance of a given game will depend on what it's trying to do across frames. How many objects is it drawing? What si the complexity of each object? How large and detailed are the textures - is there enough bandwidth to process them all quickly (well within 16ms)? How many and and at what quality will shadows/lighting be rendered? What post processing effects are needed and how large is the frame? How many and how complex are the AI/animations/physics/pathfinding/collision detection algorithms that need to be processed?
When you think about all of those tasks you cna sort of, round about, kind of brake the hardware up into two categories: The grunt power, that si the ability to process data quickly, and the bandwidth or connection to that data.
When the two are mismatched you're not going to get the best performance out fo your hardware. Imagine having an array of 1,000 GTX 780 Ti's GPU's. That is some supercomputing $hit right there. Now instead of the GDDR5 memory they would usually utilize, imagien we hook them up to a 5,400 RPM Hard drive. Lol. It would take hours to render a single frame of Crysis 3.
That system would have the grunt to render a Pixar movie frame in milliseconds, but if you don't feed it with data to process, TONS of data to process, most of those GPU will be idling, waiting for data to get to them.
That's the Xbone... except not so dramatically bad ;)
Log in to comment