The big advantage of neural compression is that it can retain the quality very well.  An example given in the paper shows us that NTC does pretty well compared to a typical block compression method (BC7), when attempting to memory-match (keep memory size as a constant.) Basically retaining much of the 4096 x 4096 uncompressed image (versus 1024 x 1024 for BC7) while using 70% of the space of the BC method. Also the compression ratio is 256 MB : 3.8 MB ~ 67:1 here. Basically NTC is getting 16 times the texels than BC7, while using less space.  

From the paper

Of course the trade-off, and why it likely doesn't have a clean packaged product implementation yet, is inference compute costs compared to BC methods. Nvidia has been steadily improving this over the years, though getting frame-time penalties lower with each iteration. 

I agree though, full neural rendering with mostly generated frames is the game changer. Though advancements like this are stepping stones to that, more or less. 

Wow you need to zoom in 10000x and put a red square to even notice. That's worth the price hike...sike

In a still image sure. But block compression artifacts become very noticeable in motion. I am also pretty sure you can tell the difference between a 1K equivalent texture and 4k one. If you ever seen one of those Cyberpunk 2077 or RDR2 "ultra realistic" videos on YouTube usually they're done by replacing the mostly 1024 x 1024  and 2048 x 2048 textures with 4096 x 4096 ones, using mods. It's noticeable, but not immensely so. 

But ultimately the point of research/experiments like this is to allow GPU's with less VRAM to get perceptively higher-quality textures, and then of course secondarily to improve compression for storage. So the point isn't that it will look a lot better in any context, but rather that it will look the same or better while using near two orders of magnitude less storage space or memory, but also (for now anyway) with a performance penalty as a tradeoff. Nvidia could have shown us an example that looked comparable to the BC7 implementation but used less space, but they chose instead to show what quality they can get using a similar OoM of space as the BC7 compressed image. 

An example of 1k -> 4k; not groundbreaking but noticeable on a large screen. 

For those here who are mostly physical and don't  buy GKC. If a game releases on card for Switch 1 and GKC for Switch 2 do you opt for the Switch 1 and then upgrade to the Switch 2? 

The Metal Gear Solid Collection Vol 2 brought this to mind. 

ala Dragon Quest VII re-imagined?  I'll purchase the Switch version on a cart of course.

Maybe I'm a dolt, but what even is the point of the Sw2 gkc version?  30fps and the reso for Dragon Warrior games is fine.

Yeah, me too. The upgrade only file is better than having to download the whole freaking thing.

To me it depends on whether I can download the Switch 2 edition upgrade pack if I own the S1 card. Didn't Square Enix for example state that their games had no update path? I have no idea about MGS Vol 2. If there's an upgrade path, I might actually buy the S1 version.

pretty sure there was a patch update when i popped DQIII 2d-hd...assuming there's one for I&II and potentially VII?  Either way, for the type of game and artstyle, any improvements over Switch version aren't a dealbreaker.