We are living in an HPC world that is abundant of hardware favouring reduced precision computing, yet scientific simulations still largely depend on 64-bit double precision computations as they have been standard for decades. We will glimpse at the question wether this is a technical requirement or rather a legacy of historical practice and explore whether structure-preserving numerical algorithms offer a distinct advantage in reduced-precision environments by leveraging their inherent stability and conservation properties to mitigate the accuracy trade-offs typically associated with lower-precision scenarios.