A bit more grounded discussion on UE5 demo: https://www.youtube.com/watch?v=5KgWJ1RxDhw
It also aligns with what I've been thinking, e.g. that the quality of the character model is still last gen. Or that these 30-million polygon models are just impossible to have in a production environment.
I've also checked the latest Megascans video:
And you know what, I thought I had a pretty beefy rig. This scene runs at 20 fps for me, in a viewport. In fullscreen mode is more like 5 fps. And it really doesn't look much better than e.g. last Uncharted games.
But right now I kind of understand why they've been pushing this "no-polygon limit" approach – so they can bind you to use Megascans and their hugely unoptimised assets. This scene is also not possible to have in a current production environment (i.e. a full game with AI logic, etc.). But AAA companies will jump at this as fast as they can, because the workflow they propose in this video is hugely beneficial to them. Once Megascans libraries are comprehensive enough, the role of "3d artist" will be to search through them, find an asset, choose from available dropdowns, push some sliders on a material level and insert asset into a level. In a sense, there will be no need for a 3d artist, as level designers could easily do that job themselves, or there will be a highly replaceable "asset finder" post, where any monkey with a week of training would do.
But that's AAA business I guess, indie games will still hire modelers and 3d artists with a classic skill set.