So a site is incorrect because it doesn't match anandtechs poor results (from low clocked ram)?
No, this polish site is incorrect, because it has a history of being incorrect and because their results does not match others at all. So either they are completely right AND everyone else is completely wrong, or it is the other way around
I think that everyone know already, witch is the most probable.
And what "low clocked ram"? They run the test at stock clocks, if they push them ram out of the recommended limits, it would be an overclock test. All in all it shows that we might reached the very end of optimalizations of x86 and x64 architecture, when Intel CPU does not show any IPC gains.
I would be much harsher on Skylake. I would try SuperPi tests first and foremost. And if you actually tried that (or know someone who did), you see that these gaming benchmarks done Skylake favor.