This is the most accurate benchmark of the ones you posted.
It's a shame that sites no longer do proper benchmarking.
Here's how it's supposed to be done:
GPU Testing:
Build a machine with the fastest CPU available. Set the game to the highest resolution with the highest settings. Run the test.
CPU Testing:
Build a machine with the fastest GPU available. Set the game to the lowest resolution and lowest settings. Run the test. This ensures absolutely no GPU bottleneck is happening. This is why tests with chips like the Pentium 4 would show 300+ frames per second in Quake 3. A new chip might get 330 frames instead of 300 frames per second. That was an accurate comparison of real performance and future proofing.
Today, we get these garbage benchmarks that are virtually meaningless. People run the CPU test at 4k resolution with the highest settings and conclude that a $50 CPU is the same as a $300 CPU because all of the chips got roughly 50fps. They completely ignore the fact that the poorly designed test was actually testing the GPU, and all of the results were the same because it was the same GPU used in every test. 2 years from now, the people who bought an i3 will be scratching their heads and wondering why they can't play games released in 2016 but the guys with an i7 can.