Mass-specific heat of water: 4.186 kJ/kg K
Mass-specific heat of air (20C, 50% RH, sea level): 1.005 kJ/kg K
_______________________________________________

Density of water: 998 kg/m^3
Density of air (same conditions): 1.2 kg/m^3
1 m^3/s = 15852 GPM
1 m^3/s = 2119 CFM
1 kg/s of water = 15852/998 = 15.86 GPM
1 kg/s of air = 2119/1.2 = 1766 CFM
_______________________________________________

Water has: (1/15.86) * (4.186) * 1 (K) = 264W heat capacity @ 1GPM per 1 degree K
Air has: (1/1766) * (1.005) * 1 (K) = .569W heat capacity @ 1CFM per 1 degree K

Said another way, you need ~464CFM of airflow to move the same amount of heat as 1GPM of water (265W @ 1 degree delta).

A perfect single 120mm rad with 0 airflow restriction (impossible) and an 80CFM (max flow rating) fan can only remove 265W at a 5.8C delta, that's the best possible scenario before physics kicks you right back in face--to get near-0 water-to-air deltas, a perfectly efficient radiator still requires way more airflow than our fans can do at a reasonable sound pressure levels. Because of this limit, we increase airflow by going parallel, i.e., multi-fan rads.

Based on skinnee's rad testing, a single Ultra Kaze 2000RPM fan (roughly 80CFM in open air) can maintain an air-to-water delta of ~12-13C at 277W. That's roughly 45-50% efficiency relative to perfect.

Radiators have two counteracting figures: air flow and air saturation. Increasing FPI and/or thickness increases saturation but decreases airflow (via restriction). We've had a few rads now that get really high air saturation numbers (high-90s in percentages), but they're restrictive to airflow. Surely rad manufacturers are working on lowering airflow restriction while maintaining saturation. But even if they get a perfect radiator, we'll only have radiators that are 2-2.5x as good as they are now, which isn't that impressive (IMO) considering that performance is already attainable by just increasing rad count or fan RPM

IMO, air-to-water deltas are impressive as is. Sure we can get more at similar noise levels with better fans and some radiator improvements--but I don't think there's much headroom there.

I still think the block is where it's at. You double block performance and you go from 76C CPU temps to 53C (at 30C water temps, 22C air temps), you double radiator performance and you go from 76C (same conditions) to 72C.