This was about the curve optimizer scaling..
There is a upto ~650mhz difference between stock (PBO) settings and CO @ -30 at the same temperature.
At -15 CO the difference is upto ~500mhz compared to stock
--> diminishing returns with curve optimizer.
Did you already know that ? I found it interesting at least.
Know what? That someone at reddit with 5900X draw a nice graph that compares his CPU at stock with PBO tuned? Then what it has to do with CO scaling? Actually I'm glad you find it interesting, but have to note that it was not about CO scaling but PBO2 tuning, where using CO at least implies substantial increase of PBO limits. Here is what my "CB R23 MT curve"

looks like, for example, for working temp. range 68-75°C:
CPU at 24/7 settings and CO is basically -15 ( its per core tuned indeed):
For "PBO auto" I choose not to run a bunch of tests, since they mostly following the common pattern. Comparing fully stock CPU with PBO2 tuned is uselees, since CPU Fmax limited to 4650 and always hit the PPT limit of 78W.
You may see some 80-100Mhz bump for 73°С for the tuned CO, 375Mhz boost override and etc. So, yes, "free" performance is nice but nothing outstanding or unexpectable.