What's changed or what breakthroughs have occurred just recently that make real-time raytracing for games feasible now, when they weren't before? Is it any algorithmic development, or is it more like NVIDIA being able to cram even more computing power in the same-sized device?
kayvonf
Many things:
Faster, parallelizable BVH build techniques
Better understanding of how to build hardware accelerators for ray-BVH traversal
More brute force compute power
DRAM sizes increasing so that scenes (and their BVH's) can now fit in memory.
Gyro
Thanks anon for sharing the videos. Seems like it'll be really hard to make a good AR glass. As we can see RTX which is specifically for real-time ray-tracing is already very expensive, so to make a good AR experience that visual and real objects can have lighting effect on each other, GPU will be a very tricky bottleneck.
mihirg
Do we use mipmapping to pre-filter the shadow map at different resolutions as well? Seems like that could help with this, right?
Some examples of new GPU technology like NVIDIA RTX that enables improved ray tracing performance:
https://www.youtube.com/watch?v=1IIiQZw_p_E
https://www.youtube.com/watch?v=WoQr0k2IA9A
What's changed or what breakthroughs have occurred just recently that make real-time raytracing for games feasible now, when they weren't before? Is it any algorithmic development, or is it more like NVIDIA being able to cram even more computing power in the same-sized device?
Many things:
Thanks anon for sharing the videos. Seems like it'll be really hard to make a good AR glass. As we can see RTX which is specifically for real-time ray-tracing is already very expensive, so to make a good AR experience that visual and real objects can have lighting effect on each other, GPU will be a very tricky bottleneck.
Do we use mipmapping to pre-filter the shadow map at different resolutions as well? Seems like that could help with this, right?