yeah like I mentioned this is an early version of DLSS upscaling, not the newer and better implementations like we see in more recent releases. What we would likely see from Nintendo's use if that ever happens, is something really focused on getting the best results from the hardware as possible. One big benefit Nintendo has to work with here is lower resolutions, with the 1:4 scaling, I assume 360p, 540p and 720p will become standard render targets, for 720p, 1080p and 1440p upscaling. That should mean less DLSS processing than something like this 4K imagine, which would have to be done via a 1080p render target, this makes sense as I wouldn't expect Nintendo to use more than 768 Cuda Cores, which current architecture would mean 96 Tensor cores, a lot less than you'd find in even a RTX 2060, so smaller resolutions should be more effective here.Doesn't FFXV use the older form of DLSS (the one that requires prior training)? Isn't the new "DLSS 2.0" as Hardware Unboxed mentioned much better utilized and actually uses much more of the Tensor cores than previous itterations?
It's also worth noting that this ability might allow 480p generation games to get auto HD renders, basically you'd be able to get a 960p widescreen image, or if the game is 3:4, you'd get the bars on the sides, but still a 1280x960 image. It's interesting to think about Nintendo's retro titles with such a technology behind it, it's also really good with prerendered backgrounds so stuff like donkey kong country could have a pretty nice upgraded look via DLSS. Sadly it's perfect for PS1 games, but Sony isn't using DLSS yet, though they could offer these upgraded games as cheap remastered classics... It's an interesting technology for low resolution consoles of the past and might be a great fit for 3DS titles as well.