Thanks to a comparison posted on the Chinese Bilibili forums, we now know just how efficient Intel’s 10nm node is compared to the dated 14nm process. The Core i5-11400H (Tiger Lake-H) which has a power draw of 52W at 3.52GHz performs roughly the same as the Core i5-11400 (Rocket Lake-S). This indicated that the 10nm Enhanced SuperFin node which powers the former is around 35% more power-efficient than the preceding 14nm node at the same frequency.

Granted, part of the performance boost also comes from the larger sub-system memory on the 10nm Tiger Lake-H part, but considering the level of maturity the 14nm process has reached, it’d be fair to say that much of those would be canceled if the latter had been around as long as the former.
According to Intel itself, the first iteration of its 10nm process is expected to be 60% more power efficient and 25% faster than the original form of the 14nm node, so this is not really a surprise. The company, however, expects the more mature 14nm+++ node to be faster than the former, but since then the 10nm node has also evolved a fair bit, adopting the SuperFin design and the MIM capacitor design along the way, these estimates do seem reasonable.