Unfortunately, AA left the realm of 'easy to solve' behind completely when deferred rendering took over and MSAA became largely non-viable.
Various solutions have been developed the years, but to my eye, all of them are messy and imperfect.
- FXAA's here, but as always, proves insufficient on its lonesome. It's "reliable", sure - unfortunately it's also reliably bad.
- TAA works well.. except for all the blurring and the need for post-proc sharpening.. and how different people can suffer greatly from its blurring.. so to make it not just a 'turn off at first sight', you've got to expose a bunch of not-always-trivial-to-explain individual settings.. and sometimes it still just won't look good. To be fair, modern implementations of it have improved, but it's so easy to get wrong.
- SMAA is cheap and efficient and never, ever seems to catch everything as well as you want it to. Also, any flaws - or "tactical choices" - in your pipeline can magnify its problems (just look at Elite Dangerous struggling with this for.. uh.. forever?)
- SSAA is not real. Okay, yes, supersampling is real, but it's not an algorithmic solution to AA, it's just pushing frames at higher resolutions and then scaling them down for your monitor. Which, like - at that point you've already paid into the ecosystem to run at the supersampled resolution, so just buy a 4k monitor and call it a day.
- DLSS is just SS"AA" with guessing, but now Nvidia exclusive and RTX required. Basically the only card that could possibly benefit from this while not having the capacity to supersample FFXIV is the 4060, and if you bought that, you already made a mistake. (Sorry.) Maybe if you're one of those lunatics that runs 200gb of upscaled world textures and are still trying to push >120fps? Still - Nvidia exclusive. Not likely to see the dev time.
- TSR is fascinating as an evolution of TAA.. and also an Epic homespun solution. Maybe you can convince Yoshi-P to switch to UE5? Don't know enough about it to give it a fair shake, but since it's never gonna show up here.. it's probably fine to skip over.
If I had to guess, my guess is that we're likely to see either no change or a backport of the FF16 TAA implementation, hopefully with more exposed settings.
(But for real - buy a 4k monitor or supersample.)



Reply With Quote




