Those that still think that Intel gives 10% generational really need to cut down on the stuff they are snorting.
It's more around 5-6%, and only when the architecture changed (eg: Haswell-> Skylake)
(9900K is -still- Skylake, even if it has another name. They simply pumped up the number of cores to 6 and then 8)
The rest of the extra performance is simply from higher base (and turbo) clocks, as well as transition from DDR3 to DDR4 which obviously resulted in less "waiting" for CPU, making it faster.
Too bad we can't pair 2700K (Sandy Bridge) with 4000 MT/s DDR4, because it would probably be almost as fast in single core as 9900K (considering that one also overclocked to 4.8-5.0... somehow!)
Why do you think that so many people are pissed at Intel for giving same old quad cores with single-digit architecture to architecture advances, since Sandy Bridge ?
And there weren't even that many architectures:
Sandy Bridge -> Haswell -> Skylake. THAT IS IT.
SB -> Ivy Bridge -> H, or H -> Broadwell * -> S, and now S -> Kaby Lake -> Coffee Lake are simply die shrinks (or "optimizations"), with very tiny tweaks, but nothing that actually boosts IPC significantly. Same cache sizes, same number of execution units... and they only did them to save money, by making the die smaller and cheaper to produce.
* Broadwell was actually a regression, it just couldn't clock as high as Haswell, so all Broadwell chips were slower than those that were "replaced". Massive fail. People still bought them... there was not much choice from the competition at the time. Bulldozer era. Ewww.
It was a very dark and terrible time in computing...
This table on Wikipedia pretty much sums it up:
View attachment 130322
Same CPU core, over and over and over again with extremely tiny changes.
Maybe Ice Lake is actually better, but it's stuck on a completely broken process... so I would ignore it for now until they get their s*it straight.