| sc94597 said: The immediate advantage I would see in providing the actual inputs would be performance. You can have a smaller, more efficient model. They probably will do that in a future iteration of it. Maybe they wanted to start out simple, release something, and then iterate. But I do think DLSS5 is a stop-gap before Nvidia moves to online learners (which they and AMD have already implemented into their PT pipelines, to an extent.) |
Yeah, I'm thinking it's current stop gap as well - it works, it's easier to implement, even if result are somewhat questionable, compared to what fully integrated pipeline approach would give, so I expect them to move to more in-engine approach in the future.
Of course, maybe their idea is that everything looks better yassified, in which case it's already doing a great job. 
