Lots of compromises were made, even for Crytek. And the point was to get it to run and look good enough, despite these compromises. It worked, but the fact is that it doesn't hold a candle to how good the full Crysis 3 was. And when someone did try to hack FG into older GPUs, they found instability and frame drops (which make perfect sense if you imagine the algorithm expecting workload to be done by a specific interval).
As it is, DLSS FG already has compromises even with Ada's compute power, which show up in interpolation errors that so many critics are already using as ammunition that FG is 'useless', 'terrible', etc. I'd say they should focus on perfecting it for Ada first before even thinking of trying to scale it down to require less compute to fit within at least Ampere's capacity.
In the example I provided for SVP, the folks developing RIFE are optimizing their models and algorithms in the hope of one day being able to do real-time High Quality Interpolation and FG for higher frame rates, and they're running into many obstacles squeezing it out of limited headroom.
I'm not sure where you're getting that everyone else with no DLSS FG is 'lame'. After all, game support is still low, and it was marketed as an new and awesome feature. If you take that as "you're lame for not having it", I think that's a 'you' problem. Everyone else will likely wait for the tech to mature before diving into it, as most have with both RT and DLSS.
