Yes and no. This whole puddlegate along with the Nvidia RTX tech has exposed how little i know about how reflections work in video games. After watching your vid specifically the section on the puddles, and reading your posts here, i have come to understand that these puddles are just textures and the reflections in them are faked. And seeing the RTX demo for BF1, it seems that is/was a norm across the board?
But i could swear seeing reflections (real-time or not, i sincerely dont know) in games like DriveClub and Infamous Second Son.
So are all rainy streets and puddles this gen just textures? If so, how was infamous able to reflect the moving car's neon lights on the streets? DriveClub looked incredible in the rain and the screenshot above is actually mine from a few years ago. those look like realtime reflections to me. Is that not the case?
I understand the confusion but most reflections in those games are the same as Spider-Man - they combine cube-maps with screen-space reflections.
SSR is what you're seeing most in those shots and it's in Spider-Man as well.
The basic gist is that screen data (stuff in screen space, if you will) is used as a source to calculate real-time reflections. They can look excellent but they have a flaw - they rely on data in screen-space. If you have something off-screen it can't be reflected, basically.
If you load up Infamous and look at something reflected, try tilting the camera down - once the object is off-screen it won't be reflected any longer. It disappears.
So cube-maps are often used alongside these to help give maintain the illusion. It's even possible to update the cube-map texture every frame to create moving, active reflections - but this is expensive and typically reserved for vehicles in a driving game (and, if you look closely, all cars using cube-maps will share one reflection source - when you drive under a bridge, that same bridge reflection will appear on other cars).
The thing is, SSR is great for a driving game with a fixed camera perspective but if you have free camera movement, it can break often.
There are other solutions too like planar reflections, which is used by games like Half-Life 2. Those aren't a good choice, though, as they are *VERY* expensive to render and have other limitations. You're basically redrawing the scene a second time but inverted (a secondary invisible camera basically provides the source).
RTX reflections avoid all of these limitations.
In the case of Spider-Man, what they did during E3 is to basically project a texture beneath the room with a rough approximation of what reflections would look like when the camera position is centered in the room. This is combined with SSR allowing dynamic objects to render properly. The cube-map is still approximate but it's hard to make out detail due to the water surface distortion. Lining up such box projected cube-maps isn't an easy task, however, and requires a lot of specific work every time it's used. I believe The Order uses a lot of very carefully placed cube-maps for its reflections but it's a super linear game.
From what I can tell, they simply eliminated those room specific cube-map textures and replaced them with ones that are more general in nature - ie, they can be utilized in many areas throughout the world in combination with SSR and still look 'good enough'.
So the E3 method isn't really more computationally expensive unless they used a super high-res texture which could eat into the memory budget but it certainly doesn't appear that way. The artists did a good job there in that demo and that's because it was a highly tuned demo - they could afford to spend that time but likely realized that it wasn't really feasible to do that across the whole game and found another solution. Just a case of running into a production problem and solving it, I'd imagine.
That's a very very basic attempt to describe it but hopefully that gives you a better idea of what you're seeing in those other games.
