All right so I came back to drop some knowledge in case anyone does a search and comes across this post. I'm no expert, so feel free to correct antyhing that seems wrong.
The noise most likely comes from the chokes on the voltage regulator modules (VRMs) on the board. The noise goes away if you (a) turn turbo mode off, locking the CPU at a lower clock speed, (b) disable idle states (aka power-saving states), driving CPU usage to 100% and having a more or less constant clock speed across the cores that is in-between base/turbo, (c) or like in the above post disable states C2 and deeper plus C1E, which will also put your clock speed at a more or less constant number like in option (b). All of which makes sense because what causes the whine is the constant variation in voltage.
The way idle states work is that they either downclock a core or reduce voltage to it, or in some cases both. It 'starts' at C0 (normal state, code is being executed, all things powered/on) and then goes through deeper power-saving states [C1, C1E...C10]. Normally with a system at idle, you should see a small percentage in the C0 state and most in the lower states (C6, C7...C10). Generally, state C1 usually works by only reducing clock speed, while the other states work by reducing voltage and stopping clocks, except for C1E (Enhanced/hybrid) which does a combination of downclocking and reducing voltage.
If you disable idle states through the Windows power plan options ("Processor idle disable"), then you effectively disable all the idle states, so you will see 100% in the C0 state and the noise will be gone because the voltage demand of the cores is not altered with time anymore through the ever-constant in and out of these power-saving states.
If you use the registry hack, you are disabling states C2 and below but retaining C1 and C1E. Now with C1E on, you will still hear the noise because as mentioned it does downclocking and voltage reduction. The minute you turn it off, you more or less have the same scenario as above except you still get access to C1 (which doesn't touch voltage, just clocks).
The downside to doing this is that you will be drawing more voltage and watts and it might increase temperatures to some degree (on my setup, voltage went from 0.79 V to 1.2 V, power from 6 W to 19 W). But also another downside is that it will affect single-core performance; the "extra voltage" released by the cores in deeper idle states is no longer there. So you will see this reflected in your Cinebench scores; for example my scores for multi-core were very similar, but for single-core, they suffered a ~30% decrease.
Also just to add, Intel's datasheet/manual does make the following mention: "Caution: Long-term reliability cannot be assured unless all the Low-Power Idle States are enabled", so take that as you will. The regedit option does not technically disable all the idle states, and this warning could just be a non-tested clause type thing but it's still there.
Another thing that helped in my case to lessen the noise was playing with some of the options in the Windows power plan options. Mainly the 'Processor energy performance preference policy (lower values favours performance, higher favours power-saving) and the 'Maximum processor frequency' and trying to cap it at different values. You can try different variations of values for both of those (also playing with the associated options for these settings that include "for Processor Power Efficiency Class 1" - I think maybe this may refers to the P-cores vs the E-cores, but I'm not sure - sometimes one and not the other), as well as still turning off just C1E through TS. All of these options didn't completely eliminate the noise, but it made it better to some varying degreese. Of course, it also affected performance in varying degrees. I also tried many of the other options for increase/decrease thresholds/policies, etc. etc, but most of them seemed to have little effect (on the noise and on the distribution to the C-states, but it's possible these were locked). It would be interesting to see what would happen if one were to turn off SpeedShift, but that doesn't seem to be an option in my case.
Bottom line? I think Asus needs to use better quality VRMs, because I saw four models of my laptop and they all had the same issue, and it is absolutely frustrating when it seems there are no reports for most other manufacturers. It is terribly disappointing that I have to tone down my CPU's capabilities just so it's bearable to use in a quiet environment.
