We are already at the point where any GPU made in the last 4 years, and any NVIDIA GPU from the last 6 years, can at minimum run raytraced reflections and shadows with the rest of the scene rasterized at 60 fps in native HD resolutions. Cube maps, SSR, stenciled shadow maps, and the like are technologies that debuted 25 years ago and have been pushed as far as they can go. Despite literally two and a half decades of iterating on them, the fundamental limitations of those methods and the associated artifacts are still obvious. At least with reflections:
- Cube maps aren't perspective-correct and are often pixelated
- Too many real-time cube maps becomes prohibitive
- Planar reflections are prohibitive for more than a couple surfaces, can't handle irregular surfaces
- SSR can't capture anything off-screen
When the min spec on games is a 20 series Geforce, which it will be in nearly all games very soon, there's no reason to use these outdated methods at all. They're a pain in the ass to implement, they're extremely limited, they're fragile, and they're not necessary to achieve good frame rates. It's time for these old methods to be put out to pasture. The only reason to keep using them is if your game has to run on the PS4.