We assess the effectiveness of the temporal fence instruction, introduced by Wistoff et al. (2023), primarily in enhancing both security and timing predictability for embedded systems. Originally demonstrated to eliminate microar-chitectural side-channels by flushing the on-core microarchitectural state using the seL4 kernel, we extend the evaluation of the temporal fence to a minimal kernel using RISe- V's PMP to protect applications, examining its implications for both security and safety. Our experiments demonstrate that by using scratchpad memory and the temporal fence, we can improve the security and time predictability of applications on both embedded and application-level processors. Furthermore, we find that the performance overhead introduced by the temporal fence remains within acceptable limits for hard real-time systems, which demonstrates the temporal fence's potential in securing applications against side-channel attacks while enhancing system reliability.