You have missed some important developments in the meantime. CPU overclocking is all but dead nowadays.
For Intel, it started with the transition from Coffee Lake (Core i-8000) to Coffee Lake Refresh (Core i-9000). There, they pushed the CPUs quite near to their maximum by default. While the 8700K for example was one of the best CPUs to overclock in years, the 9900K was almost not overclockable. The different turbo modes had the CPU running at the limit of the voltage/frequency curve already. With any overclock, you would therefore get a massive rise in power consumption almost immediately, for very little performance gains.
With the Core i-10000 series (Comet Lake), it became more extreme. The 10900K is so close to the limit that any OC equates to squeezing water from a stone. Depending on the cooling, often times performance even decreases with OC, because the three (!) different intelligent turbo modes can't work properly or at all anymore.
Since that time, overclocking with Intel consists of removing the PL1/PL2 long/short power limits in the BIOS, and hoping that your mainboard's VRM section and your CPU cooling can cope with it. Strictly speaking, that's not even overclocking, but it's the best way to get "free" performance from the CPU. Generally, it's much more fruitful to OC your RAM instead.
With the upcoming Core i-11000 series (Rocket Lake), it will get more extreme again. Like stated in the Anandtech preview, we're close to the limit of conventional cooling methods, regarding hot spots on the CPU die. The Z590 boards had to upgrade their VRM sections again, and therefore become more expensive again, if the top CPU models should work flawlessly. Because good
powerstages can easily cost 10x of what seperate Lo- and Hi-MOSFETs cost, and you have 16 and more of them on the higher-up board models. Therefore they're spending way over US $100 on the VRM section alone, so the platform cost is constantly rising since at least two generations.
With AMD, it's much the same story, since Ryzen 1000, but even more so since the Ryzen 2000 series came out in 2018. The Precision Boost 2 takes those CPUs right to their limit already. That's why Intel had to "counter" with the Core i-9000 series also not leaving much in the tank regarding OC headroom.
Many CPUs from the last two years are predominantly most efficient at stock settings, and dramatically lose efficiency once you above that. Even PBO (Precision Boost Overdrive, a sort of "official" overclocking that AMD offers), already causes that. With PBO, the clocks will be up to 200 MHz higher, which translates into maybe 3% performance improvement; however, power consumption and thus heat production already increase considerably. Thus, even PBO is not advised, let alone overclocking.
Here, you can see how a lower PPT limit (Package Power Tracking, forcing the CPU into another mapping of VCore vs. Frequency) can improve the efficiency per task, which is quite telling:
The lower PPT limit will quickly improve efficiency (power draw is decreasing much more than performance is), since AMD went so close to the limit already with the latest CPUs (just like Intel). Undervolting with an offset isn't nearly as effective. On the other hand, overclocking manually is a bad idea too. It cancels out the fully intelligent power management and the finely-tuned turbo modes. The CPU at stock, without PBO and manual OC, is already slightly above the most efficient range. PBO will push it over it, and manual OC will make the CPU massively inefficient and a [***CENSORED***] to keep cool. Going above the stock settings with any current CPU, especially the higher models, is like having water boiling nicely in a pot, but still turning up the stove because you want the water to "boil more".
That's why you should concentrate on RAM OC, much more benefits to be had, without any of the penalties in efficiency.
