I think it is something within reach for Luxcore renderer.
Let's say we manage to get something Like path guiding speed bump onto GPU resulting in a 6X to 8X speedup.
Then Let's say Upscaling become available from Nvidia or AMD or INTEL So here we can get a 4X time speed up using half the original resolution
Then you have Something like Nvidia Slow motion AI to render less frame https://youtu.be/MjViy6kyiqs
as seen here can result in a 3X speed up.
Let's do the Math : 6X4 = 24 and 24X3 = 72X time speed up. Now let's you have a project where each frame take 20 minute on you 2060S to render without Cache and apply a 72X speed up on it :
20mn in second = 1200 sec
1200 / 72 = 16,66 sec per frame.
Let go even further you drop the 2060S for a 3080 you get another 4X speed up : so from 20minute to 16.66 sec now to 5 sec per frames.
Now is all that Realistic ?
Let's take actual solution:
Never seen something like that for GPU but there is all that on CPU (and it is where the 8X speeed up came from).But we don't have it now.
Aside from DLSS wich work on raster and DirectX there Ai Gigapixel solution wich work pretty well on image sequence.
Better there is davinci resolve superscale solution but it actually in the non free version
Never see a Nvidia SDK for Devs about Slow mo but there is Davinci resolve Solution for free : https://youtu.be/GkKdKi6ZLUI
Yes the 3080 do exist even if it is hard to pick one atm but it is a question of time. And you can even take 2 of them.
But it is a bunch of IF