Man those RT reflections look great. I hope I can enable them without FPS tanking too much (with DLSS). The rest of the RT features I can live without.
Ideally, yes, as that would be how RT functions. Rather than creating a separate shader for "mirrors" and "puddles", etc., you can simply treat them like any other surface in the game. The engine will simply be told to reflect more like 100% of the light that strikes it with minimal refraction and no alteration of colors. Hence, that surface will look like a flat puddle reflecting all of the world around it. If I apply a red texture, and tell the engine to reflect only 75% of light that strikes it, I could make that same puddle look like a pool of blood instead: duller, deep red, and showing only muted highlights to make appear wet and shiny. If I applied a white texture that looked like a pattern of cracks and told the engine to only reflect 50% of the light that strikes it, I could make it look like ice: more matte, and the texture applied would give the appearance of cracked ice.
The more variety and complexity I add to a scene, the slower the performance is going to be. However, the days of needing to "artificially create the illusion of reflections" by loading shaders that needed to actually invert the vertices of 3D geometry, their textures, etc. and then render and draw them onto the "reflective surface draw space" as a separate pass before rendering / drawing the composite of both frames during a third pass in order to produce the final frame...this will be a thing of the past.
Ray tracing is using the same data to create lighting effects regardless of whether it's supposed to look like wood, blacktop, shiny leather, skin, or chrome. To alter the appearance of everything in the scene if I were to, for example, shine a flashlight on it, it doesn't require me to create "glow shaders" to "paint" that extra lighting over the existing textures. I just create a source of light, and it will add those rays into the scene along with all of the other rays that were already there -- interacting with all surfaces and all other light sources equally.
The downside is that this is a bleep-ton of processing to do overall, as ALL of those rays need to be
constantly updated, "pixel-by-pixel", across every single frame. It's a fantastic idea because it's a universal system, not requiring different techniques for different effects. It's just something that neither software nor hardware has been capable of doing up to this point. So, it's a bit of a
future blessing:
As hardware becomes more powerful, rather than adding even more layers of rasterization to achieve more detailed lighting results requiring more and more rendering passes...we can now devote 100% of the processing power to a
single lighting technique, managing most effects in one pass. Hence, right now, hardware will struggle to do it -- but 5 years from now, it will be MUCH faster and more efficient to use RT instead of additional layers of rasterization to achieve the equivalent levels of detail.
