Their "EFPS" testing methodology seems rather questionable, as they are only focused on a handful of older and/or otherwise less demanding games, running at 1080p resolution. This testing won't show the limitations of a 9600K's 6-threads, nor will it show how in graphically demanding games, especially at resolutions above 1080p, the graphics hardware will be what's limiting performance more often than not. In these less-demanding games, even the 9600K is technically "bottlenecking" the 2070 SUPER's performance, just to a bit lesser extent. If one is only interesting in playing these games at this resolution, they probably won't see much benefit from a $500+ graphics card over one costing significantly less.
If they really wanted to provide an accurate comparison of hardware performance, they would be testing more than just a handful of esports titles at a lower resolution and calling it a day. Instead, they should also be benchmarking recent releases at a variety of resolutions to provide a better picture of how the hardware will perform in current and upcoming games.
There's also a lot of sensationalist writing on that page, and in some ways they are being downright manipulative. They open by claiming that a 3700X gets less gaming performance than a 9600K while costing more, but fail to to mention that both processors are overclocked in their test, and that the 9600K requires decent aftermarket cooling to get that performance, whereas the 3700X gets practically its full performance without overclocking on its included stock cooler. In fact, there was no need to even make it an AMD vs Intel comparison, as an overclocked 9600K should perform practically the same as an overclocked 9700K or 9900K in that limited set of lightly-threaded benchmarks they ran, apparently making those processors redundant as well.
I do visit UserBenchmark a fair amount, but that "EFPS" testing seems like complete garbage in its current form, about on par with other useless "bottlenecking" sites that only base their results on a very limited set of data that probably won't apply to most users.