curl-6 said:
sc94597 said:
So the two types of DLSS have been confirmed. Given this, I suspect by the end of this generation we'll have many more DLSS versions, including some that cost as much as type II, but have the quality characteristics of type I or even better. Models are improving over time given the same parameter count, and not just because of architectural changes, although you could ostensibly transfer learn from a ViT to a CNN to improve it, and likewise from large models to smaller ones. As we also get more RT titles, I would like to see some ray reconstruction, even if it is the older lower-cost preview version. That's probably the most exciting part about SW2. Nvidia's evolving feature-set, which has very much been backward compatible, benefits it over time akin to the old "optimizations" that consoles would experience over the course of a generation. My guess is a SW3 will go hard on neural rendering if it releases in 2031-2033. |
Interesting stuff. I feel like we should see some significant graphical growth for Switch 2 over the next few years, not just cos of this, but also the fact that third party devs apparently didn't even have the final specs for much of the development of its early titles, and that Nintendo's own games so far have been pretty much all been basically Switch 1 games on steroids. |
Honestly, my take away from this is that devs who use DLSS should stick to outputing at 1080p and use "full" DLSS model, cause that "tiny" model really breaks apart - as lot of people were noticing from the start, with Hogwarts being first suspect from the very first trailer.
