VXGI is the direction that Unreal is heading, which is fine because I’m assuming they worked a deal out with Nvidia. Both SVOGI and VXGI use similar methods, the latter just uses 3d clipmaps instead of octrees. The pros/cons between the two methods vary. 3d clipmaps(VXGI) will give you better performance, memory usage, and handle dynamic objects faster, but you will end up with varying density aka varying quality(stuff in the distance won’t be as accurate). They just need to finish working out some of the kinks to it involving tracing/blending to prevent flickering due to the changing densities in the clipmap. I’m assuming that’s where a lot of the performance hit resides at the moment.
Here’s an extremely good writeup comparing the two: https://erkaman.github.io/img/master…thesis_101.pdf
If you look toward the end, you can see some performance comparisons.
And I’d definitely agree with you on the landscape part…