I'll bet dollars to doughnuts that even twenty years from now, by far the cheapest way of creating a completely photorealistic VR river scene is to build the actual river and record video of it. Then you "only" run into the problem of storing and distributing the petabytes of data you record.
But... look at the ground and you'll just see a flat greenish noise texture. They try to cover it with clumps of grass and vegetation, which works to a degree, but not completely.
Look at the water flowing - particularly on any kind of whitewater rapids/waterfall - compared to the character textures and the buildings, it's really quite disappointing.
You need a fairly complete simulation to really fool the human eye and we're a long way from that.
As for a "VR" river, there's nothing wrong with filming a real river with a 360 degree camera, and then augment that scene, rather than "building a river". A real river on a virtual Mars, might look better than a virtual river on a virtual world.
Of course you can dream up a game that hard to bring to life exactly as you envision it using current technology. That doesn't mean new technology doesn't open up a few new possibilities. If you run into problems thinking big and complicated, try thinking smaller and simpler. I'm sure a Populus style game would be a lot of fun with the DK2 (isometric sandbox god simulator game with slightly cartoony graphics). Why does it have to be realistic?
[0] - keeping out the "real reality" from showing up involves obvious points like keeping distractions (especially other people) away, but also some meta-points; if your product has glitches or even simply keeps "breaking the fourth wall", it may be just enough to prevent any kind of immersion from happening
10 years ago, the state of the art was something like Nvidia 8800 gtx - it looks like the recently launched 1080 is a handwavy 4-10x in terms of performance[b]. ~10x the RAM, but not bandwidth etc. Having recently(ish) upgraded from early 2000s system to an i7 with an AMD R9 -- I would anecdotally claim that we've come quite a long ways in ~10 years. Even if part of that is "just" making 2k-4k gaming a reality.
I think it is a little bit like optimization of performance. Small changes in aliasing, changes in resolution, refresh rate -- they all add up to move us closer to "the real". And I think one shouldn't ignore the work done on the modelling, skeleton, animation and physics side either. It all helps make artificial worlds feel "more real".
With all that, I don't really see why "realistic" is such an interesting goal. For simulation (be that virtual tourism, or a driving simulator helping people become better real world drivers) - sure. The tourist want beautiful scenery, the driver wants real physics (and neither would mind getting both).
But for gaming -- chess isn't very realistic, go is rather abstract and tabletop roleplaying games have the best graphics engine - shared with literature - the human imagination.
https://www.game-debate.com/gpu/index.php?gid=1701&gid2=879&...
I feel like I've been hearing this statement for decades now. The human brain is remarkably good at pattern separation, and given enough time and exposure to a stimulus it will work out what is real and what isn't. Until we have a simulated reality that is as information rich as reality we'll always be able to tell the two apart