UEFI in and of itself didn't do anything to improve OCing and early UEFI implementations weren't very pleasant to work with, in my experience.
The era of jumper OCs was probably the easiest, mostly by virtue of having the fewest meaningful options to tune and test. Before that OCing involved swapping out oscillators with a soldering iron, and after that the number of settings to configure ballooned.
I know.
If the OP doesn't have a the replacement CPU yet, they should shoot for a 3770K, if practical, even if they aren't OCing. It's clocked significantly higher out of the box and has the potential to be OCed appreciably if the OP decides to dabble in that.
2. random search for 10th gen overclock:
3% increase in performance
Overclock Your intel i9-10900K to 5.1 GHz with the MEG Z490 ACE
Intel processors have offered excellent overclocking headroom for years, and with the 10th Generation CPUs, this hasn’t changed one bit! Fine-tuning your performance will net you an excellent performance gain with minimal-to-no risk to your hardware.www.msi.com
Comet Lake is not remotely comparable to Ivy Bridge in terms of overclocking potential.
Intel's fabrication issues and resurgent competition from AMD for the several years has caused them to bin and clock their parts much more aggressively than in the past. The 10900K is the fastest part of it's architectural generation and the highest clocked part ever made on Intel's 14nm+++ process. It has almost no headroom, as that MSI article shows.
The 3770K, while occupying a similar segment, was many generations earlier, during a period where there was almost no competition from AMD. A typical 3770K sample is good for ~15% more performance from OCing, without needing much in the way of fancy support hardware.
Some other parts can OC much more. For example, in semi-recent memory, I had more than one Pentium E2140 that took 100%+ OCs, with a corresponding increase in CPU performance, on $70 motherboards, completely stable on surplus Intel boxed coolers for a slightly higher-end Core 2 parts. All of my Xeon X33xx, i7-920, 970s, X56xx, 2600K, 3930K, 4930K, 5820K, and 6800K samples all took 20% or greater (often much greater) OCs, as did most of my later A64s, Opterons, and mid-range FX parts.
Even for newer parts with negligible core clock headroom, there are often gains to be had from overclocking uncore, memory controllers, various interconnects, and system memory. My modern (AM4) systems, with generally marginal room for manual CPU OCing, are still noticeably slower if you pop into the firmware setup and load 'optimized defaults', or tell them to just use XMP settings for the memory.
Even the non-K parts can have their top turbo multiplier applied to all cores on many boards. Of course, this is not going to help apps bottlenecked by only one or two intensive threads.
