We have seen a lot of great archviz scenes with Vray quality in Unreal, but actually a game with the same or nearby quality is impossible for obvious reasons, a medium house with extreme lightmass values to reach Vray quality runs in 60 or even 30fps in a GPU like GTX 970, maybe a interactive experience in a house is possible, but not a complex game like Bioshock, Half Life, Fallout etc. Soon 4.11 is being officially released with big improvements in lightmass, and will come a promissing tool, which are Light Portals, and the results looks very promissing, with him is possible to reach a good quality with a low performance impact. Does now we can make something? What you guys think?
And sorry for my bad english
Yes, but it depends on the game. You could easily make something like Portal or even a small arena shooter map like Unreal Tournament with archviz level graphics.
Right now most arch-viz ‘‘scenes’’ are stripped of everything that’s needed in a game. Complex game mechanics, A.I, characters, animations, vfx… etc. So it’s easy to focus all power to the shaders/lighting, even if the fps are low it doesn’t matter too much. But for making a fully cohesive game with high framerate you always need to make sacrifices. That being said, games are becoming more realistic everyday! It’s always the hardware that’s lagging behind hehe!
You’d need to forget about dynamic lighting though.
Just a quick note, with lightmass you won’t see much difference in FPS on the gameplay machine, but you will on the development machine which the light is building on. Light is precomputed and stored in light maps on your development machine(s), and the machine you actually play the game on, I’d presume, pretty much just multiplies the surface by the light map to apply lighting (I presume this is what happens, not sure on the details). So if you crank the Lightmass settings way up, there’s absolutely no difference in FPS between Lightmass on lowest settings and Lightmass on highest settings, on the gameplay machine. But, if you crank the settings up, lighting build times will rise very quickly. That’s where the performance payment is made, on your development machine(s) whilst building lighting, not your gameplay machine.
On to the actual question, the answer really depends on whether you’re talking static or dynamic lighting. Say in a complex game environment like GTA V or The Division, where there’s heaps of surfaces to be computed. With static lighting, light build times will be incredibly long, but gameplay machines can play without much problem, and you’d have incredibly realistic lighting. At the cost of lighting being static, mind you, so no moving light sources and no moving geometry. So with static lighting, again, arch viz quality in actual games can be done but is costly in terms of build time. Dynamic lighting, is different. With dynamic lighting, you’re relying on dynamic solutions which are orders of magnitude less realistic than arch viz tends to be, so there’s one factor. Another is whether or not you’ll be using GI, note that GI can quickly make or break a scene in terms of realism. Say you’re using VXGI, I’d say that VXGI would have issues with such a complex environment, even on high-end cards. And you also have to factor in soft shadows, which can be done in several different ways (PCSS / CHS IIRC for NVIDIA and AMD respectively, or distance field soft shadowing within Unreal for lights that support it). So the answer for this is as follows. Static lighting? Sure, you can reach arch viz level lighting, if you’re willing to pay the price in build times and memory requirements for building, which is so ridiculous that static lighting isn’t an option for basically any open world games unless you make sacrifices that reduce the quality. Dynamic lighting? You’ll have issues getting quality like what’s shown in arch viz scenes, not only for performance but just lack of techniques for certain things.