fatslob-:O said:
Culling and compression is pretty much as good as it gets. There's alot of games that use extreme forms of culling such as occlusion culling and compression ratios go as far as 4:1 for 32 bit colour space. Last I heard rasterization is waaaaaay more efficient in comparison to ray casting or ray tracing and other monte carlo methods to rendering.
|
Nah. You can improve the compression if you're willing to spend compute cycles, 8:1 and 16:1 can be achieved, not only that but more data could be compressed too.
AFAIK only Textures are compressed on the PS3, you could also compress things like Normal Maps, light maps, shadow maps, HDR Textures etc'.
That would conserve allot of memory, then you can drive up the quality of each.
Plus, they could "bake" more things into a scene, that don't require extra passes.
Also, there is more to rendering than Rasterization, Ray Tracing and other methods are pretty demanding with the pinicle probably being Photon Mapping.
nVidia however did attempt to not "follow the crowd" during the "3D Revolution" where they attempted to use Quadratic texture mapping rather than the standard Polygon with the NV1 core.
As for culling... It can always be improved.
The Z-buffer isn't very smart and there is always at-least a little bit of overdraw, sometimes by a fairly large amount as the engine attempts to predict where the user is going to be looking next.
Besides, developers have for a long time now been playing a game of "Sacrifices" for example Bungie removed the Tessellated Water and Triple Buffering from Halo 3 to free up resources, throw in improved compression and impostering and that allows Reach to offer better lighting, textures and polygon counts and a much much much longer draw distance. (Mostly thanks to impostering where it replaces a 3D model with a 2D sprite.)
Halo 4 took it a step farther and went with simple geometry for the games architecture and baked shadows in the game world to bolster character models and lighting substantually.
Essentially, each iteration looks much better than the prior, of course you would get some low level optimisations thrown in too.
There is still more to be squeezed out of the hardware though, well. In my opinion anyway, but I wouldn't expect anything massively dramatic.
Licence said: Well, he is right, of course, but at some point it becomes cheaper to just let more powerful hardware do the work than to optimize the software. As a simple, entirely contrived example: If you have a program that needs 100Mb of RAM, but the PC you are running it on only has 50Mb of RAM, the question is: Is it cheaper to redesign the program to work with 50Mb of RAM, or is it cheaper to just add 50Mb more RAM to your PC? Another reason why better hardware do not necessarily translate into better software - the developers may not make the same investment in optimization on better hardware than they would've done previously. |
You only have to optimise the game engine to do "more" only once, then you can use it for all your future games, depending on the game engine, that could be dozens of titles.
As for the PC, you have the option of doing neither, just let it fall back to virtual memory. :P