Yeah - Apple can't escape the laws of thermodynamics: you want more top-end performance on the same node size --> you need to up the frequency, pump in more power, accept the hotter SoC and non-linear returns.
Once the node size shrinks next cycle Apple again, in principle, gets to choose either increased performance or more efficient chips.
That said, Apple did improve efficiency of the CPU package on the M2 vs the M1: data from Max Tech shows that when constrained to around 10 watts, the M2 performs some 10-15% better than the M1 on same watts. Also when running non-max. scenarios, the M2's CPU package runs cooler overall.
Notebookcheck's article also gave us good insights into throttled performance (M2 > M1), comparable surface temperatures, watts at idle etc., even if their main conclusion was pretty Captain Obvious-material 😀