As games become more and more threaded, the CPU bottlenecks are shifting from core clock to RAM speed. Here is a test I ran on the Gears of War 4 benchmark several years ago.
Gears 4 (a DX12 title) scales uniformly with RAM speed. This implies that any CPU bottleneck is tied to the RAM speed, the CPU is waiting for new data to be loaded from RAM. In this kind of bottleneck, overclocking the CPU itself would accomplish nothing.
What is interesting, is the data from 32 GB of RAM. With each channel populated with two RAM banks, a saw a 10% uplift in performance. That performance gain was equal to going from 2666 MHz CL16 to 3200 MHz CL14 (also 10%).
The benefits of interleaving on a channel (dual rank stick or two DIMMs per channel) seem to outweigh one DIMM per channel sink rank configurations as long as you don't have to drop clock speed and timings substantially. To that end, if the 32 GB CL16 setup you are looking at runs at a similar speed to the 16 GB CL14, you will likely see better performance with the 32 GB setup.
