Summary:
Many games see noticeable improvements, but how much of an improvement will vary. Games that are bottlenecked by GPU or memory bandwidth benefit significantly, whereas CPU-bound titles only see small improvements.
Arkham Knight, famously one of the Switch’s worst ports, is now a playable 30fps. Dragon Quest Builders 2 is… playable but still not great, building as much as possible to stress test the hardware can drop to single digit framerates on Switch 1, that’s now around ~20-22fps here. These are the two most demanding titles tested, which means that most everything else came out pretty good.
The obvious caveat here is that games cannot exceed hardcoded targets. Games with uncapped framerates and dynamic resolution will be able to take advantage, but capped framerates and fixed resolutions must remain so.
I didn’t expect it would enhance framerate at all without a game patch to be honest.
I was expecting something like new 3DS, where all games that were not specifically patched for it ran exactly the same. But I guess the difference is that new 3DS must have run in pure hardware old 3DS mode for those.
I felt DQB2 was already somewhat playable, but I probably never did very crazy builds. I remember people warning that destroying mountains on the main island for example was a very bad idea, because they were supposed to limit how much of the island the game had to render. Maybe I should check my old island on Switch 2.
N3DS downclocks to O3DS speeds on any game not specifically tagged as N3DS-enhanced. Softmodding can enable full clock speeds though.