Only for multi CPU mobos (and that would be pinning a thread to a CPU/core with NUMA enabled where a task accessed local ram instead of all system ram). Even then, I think all ram would run at the lowest frequency.
I've never mixed CPUs and RAM speeds. I've only ever worked on systems with matching CPUs and ram modules.

I think the hardware cost and software complexity to achieve this is beyond the cost of "more ram" or "faster storage (for faster swap)"