If I want to optimize for area, does it make sense to push the sampling clock the higher end?

Assuming SQNR is sufficient, and output sample rate is fixed. If I double Fs, I can shrink integrator Cap by 2x.

Other performances:

- Noise is not changing as R/gm doesn't really change
- OSR is 2x but little impact on perf as Q noise is small
- excess delay can be compensated

any other item that I missed? Thanks.