Seems to me that the graphics in games has only had marginal increases in quality for the past year or so. There are some good looking games out there, but when can we expect the next big jump in graphics technology, similar to the time when Far Cry, Half Life 2 and Doom 3 came out?
After the next console gen comes, probably. Also, after AVX2 becomes mainstream. Only in the last year or two have we finally gotten games with lighting of Id Tech 4 or better quality, FI, that don't use the Doom 3 engine. Not only that, but there have been steps backward, too, from even Id themselves . While others have met their quality, added geometry, and improved performance scaling, Doom 3 (Id tech 4) and Crysis are still about as good as anything out there.
Some games are now making good use extra CPU cores, and physics is really improving, after stagnating for awhile. We could get some real boosts with DX11 as a baseline, but major improvements will rely on artists who understand the technology,
after more engine features are well-implemented. Big game companies that treat their content guys like their code monkeys are probably going to keep doing the same old thing.
We are still very much limited by what our CPUs can do, and what content makers are willing to put effort into (edit: also content creation application creators). Clipping, flickering, bad decal implementations, bugs with baked-in lighting, etc., still permeate games. IMO, the next big jump will include a reduced reliance on 3rd-party content creation applications, especially as it concerns models and animation. Make initial high-LOD models outside of it, but then everything else should be done with an engine-specific content creation kit, that renders with the game engine itself, and has many specific tweaks for building a playable world. Turning something that looks awesome in 3DS Max into something that looks OK for common GPUs is the wrong way to
think about it, regardless of whether you use such a tool or not. I think Crytek, FI, is on the right track, especially offering an SDK as freeware.
In terms of graphics detail, we are already well into diminishing returns. In terms of immersive game feel, we've only
just improved over the early Havok games (IE, going from added-on gimmick to an integral feature of the game world).
all the "graphic improvements" you have been seeing are just shortcuts to make games look better.
That's been the case for a long time. Quality lighting and shadow are the only real exceptions, and they are themselves full of little shortcuts. That's why raster can keep on chugging, despite all the ray-tracing fans .
If Intel wanted too they could probably get Havoc to do that. But their IGPs are not the best for OpenCL.
But, their CPUs
are, and Haswell will finally bring some major vector and multithreading improvements, shortly followed by AMD, no doubt.
The next breakthrough in graphics will not happen due to more beefy hardware but through the industry finding and embracing ways to make content generation easier reducing the load on artists. It takes around two weeks to make a single high quality 3d model for a single person working on a AAA game iirc. I bet we'll see more emphasis on procedural generation.
:thumbsup: We need that to be quicker/easier,
and we need to be able to build creatures. Define skeleton, muscle, fat, skin, random filler tissue, etc.
(including being able to define custom material properties, for robots, anatomically incorrect critters, physics-defying being, and whatnot), as part of a model, have the engine's SDK process it, and then work from there.