again that would be the post process effects and heres the kicker the hardware available CAN HANDLE IT NO ISSUE its CODING thats still the problem,
dynamic lighting is doable Crysis did it 5 years ago, interesting to note no one uses it now, its still faked lighting with bloom effects etc,
ambient occlusion isnt even neccesary its used because lighting in game engines is imprecise it dosent bounce around realistically thus it cant always shadow things properly thus AO is used to forcibly darken areas that need it this points back to lighting
thats why games dont look like the movies, we can push polygons where the need to be but old dirty code holds back the whole show,
Gamebyro based on Net Immerse code is ancient pre 2000 to late 2001, since then its had minor updates for Direct X but no major changes its cpu code is based on old x87 ancient code used in the 8086 days, its getting updated but updated code isnt new code, new streamlined code is needed,
COD another example of old code aka Quake 3 1999 based
the only game engine ive seen change the way things are done is the 4A tech engine used in Metro 2033, and even then its still heavily borrowing on X-Ray engine, but its getting close,
aka they stopped using threaded code and build the CPU backend to look at things like tasks, each physics calculation each animation is a seperate task that gets accomplished thus why CPU dosent matter so much, long as it meets the minimum theres seldom an issue, GPU code wise its terrible mess look at Direct Compute its used in a way that its not ment to be, slows up the whole show,
but to give you an idea:
4A tech engine performs on console 3000 tasks per 30ms Frame,
considering a PC CPU from the socket 775 days aka Core 2 Duo is capable of far more, even the minimum cpu today can handle all demands that game engine has, with quadcores and 6 core cpus increasing that nearly exponentially in terms of what i can handle vs what it needs.
again if we look at the way GTX 580 and 7970 are now if you used all those shaders to handle texture and post process, but force the tessellation engines to handle ALL geometry then visuals could be brute forced at this point, now a game engine that could do that would take around 3 years of coding, that means 2015 the 580 and 7970 will be entry lvl, let that sit in your mind a minute,
we already have the means for it now, just all code is pretty much worthless its old stagnant etc, thats the part holding us back, i can already render 45-70 million polygons in real time when sculpting in mudbox with Ambient Occlusion On etc thats running on a single 6950 and i hit a vram wall not a GPU one,
simply put developers can go nuts on polycount if they want right now it wouldnt hurt in the least. as what we actually need for good graphics in terms of polycount and what gets used tessellation wise are two very different things. just need tessellation engines to replace regular rasterization methods.
essentially it all comes down to something simple a saying my 3D Animation Instructor told me "Its not about getting something perfect that makes it real, its about making it imperfect"
rendering does everything perfect models are constructed from polygons that are math wise perfect, boxes are perfectly flat we know in real life this isnt so. Look at characters in most games faces are perfectly symmetrical in reality this isnt the case, its all those little things that add up
I know im rambling and most wont understand what I am saying do to my inability to put periods anywhere.... but fuck it,
bottom line is we have the hardware today, to do what we see in movies in realtime, ancient unoptimized code inefficient APIs are the things hold us back,
