I doubt that number, too. Our CPU-bound workload runs "only" 20% faster on an i7-6xxx than it did on an i7-920 (calculated for equal clocks; due to the higher clock it was actually about 60% faster, which caused me to do the math because I feared a regression).
That's far away from 20% per generation.
But in one thing you err: After a small custom BIOS mod, my home PC now boots from nvme - with a Z68 board (Sandy Bridge era), the module sitting in a PCIe adapter card. Was just a matter of adding some UEFI module.
Yeah, my mb had an nvme m.2 slot, but it was pciex2, so the x4 card (most seem to be) didn't work. Why I went with an adapter.
The main thing lately is when I'm running a lot of containers/vms for workloads it tends to box down a little. Outside of that and 4k 3d, don't notice it ever. But the new stuff is only just starting to get better enough that I may upgrade next year sometime.
M2 SSD is available on Haswell (Xeons at least) via a PCIe Board. Thats more than 4 years old.
Source? I suspect you're being shown a few choice benchmarks which demonstrate some improvement, perhaps where something became newly vectorizable...but realistic and consistent benchmarking doesn't show nearly that level of improvement.
One area in which you may benefit is in peripherals - an M.2 SSD and DDR4 RAM will help, and those aren't available on 5 year old motherboards. Also, power savings may be found with newer systems if you're in a laptop form factor where that matters.
But let's not pretend that single-threaded performance has doubled in 5 years. Here's one article that goes back quite a ways....and 2012 SPECint comes in at about 60,000 while 2017 comes in at 83,600.
https://www.karlrupp.net/2018/02/42-years-of-microprocessor-...