Posts: 3,718 +4,814
Yeah, that's what people don't understand. Nobody is saying that ray-tracing is bad or stupid because the concept is amazing. The problem is that nVidia pushed it onto the market before it was ready (when did the RTX 20-series come out again?) and with current hardware, it's both stupid and bad to implement it. I mean, unless you're rocking a $1,600 RTX 4090 for gaming (in which case you have a whole other set of problems to address).RT will become EVERYTHING (full replacement for standard rasterization) in the future. Far far future. When video cards are 1000 times faster than they are now. And then many many years later, physics will become everything. We won't have separate graphics cards, audio chips and physics chips. Everything will be simulated by the physics chips.
That's some great insight Hodor and I completely agree with you!But until then, the only real and noticeable use of RT are shadows in dynamically generated worlds. To me it's irrelevant whether shadows move a tiny bit when the Sun moves a tiny bit. But when you generate new objects in real time (in procedurally generated worlds) there are no shadows at all. The game needs to generate them from scratch, in real-time. For every freaking object. Because there are no precomputed lightmaps for objects that didn't exist at design time. So real-time shadows and ambient occlusion must be really fast. And that's where hardware supported raytracing can really speed things up.
As opposed to procedural worlds, in worlds made entirely by designers there's no huge need for real-time shadows, as shadows can be precomputed and stored to textures. Yes, they are static and won't follow the Sun movement, so just fix the Sun position and problem solved. Or create separate lightmaps for morning, day, evening and night.