Premise

Investigate how much of a difference higher clocked memory actually makes for games on the Skylake platform using DDR4.

Results

Source

Hardware Canucks tells us a different story, within their testing they use a i7-6700k

Legit Reviews's results are even smaller (very low single digit %)

Hardocp 12.2% and 10.6% differences respectivley going from 2133 - 3600 mhz

Hardware Unboxed

Discussion

Digital foundry:

• Overclocking the i5-6500 here brings some interesting gains however we're primarily looking at different clock rates of RAM on its own (data is being skewed).
• Latency between the modules is not mentioned at all. To find these results digital foundry has over/underclocked their RAM.
• Removing the overclocked results reduces the range of results.
• RAM clock speeds has a higher impact to the range of results for Average FPS (14, 13.3) than to the minimum results however we see a higher differences for some games like the Witcher 3.

Hardware Canucks:

• Only using a 6700k.
• Much larger range of of RAM clock's and has latency mentioned.
• Two games showing minimal differences.
• 7.1% difference between DDR3-1600 to DDR4-3200 for X3: Terran Conflict and 1.2% for Final Fantasy XIV: Heavensward Benchmark (a much larger range that what digital foundry used).

Legit Reviews:

• Almost flat results.
• Only using a 6700k.
• Large range of clock speeds tested.

Hardocp:

• Tries to remove the GPU from the equation
• Shows CPU ocing skews the results
• Shows Haswell performing better with 1866mhz ram!!!
• Shows a 12.2% difference between 2133MHz and 3600MHz on Bioshock Infinite. 10.6% for Metro Last Light.

Hardware Unboxed:

• Most results taper off at 2133MHz or 2400MHz.

Conclusion

As long as you have at least 2133MHz RAM you're fine, however it can be worth getting higher clocked memory but you should only consider this when you're already CPU overclocking.