Designing mixed criticality real-time systems raises numerous challenges. In particular, reducing their energy consumption while enforcing their schedulability is yet an open research topic. To address this issue, our approach exploits the ability of tasks with low-criticality levels to cope with deadline misses. On multiprocessor systems, our scheduling algorithm handles tasks with high-criticality levels such that no deadline is missed. For tasks with low-criticality levels, it finds an appropriate trade-off between the number of missed deadlines and their energy consumption. Indeed, tasks usually do not use all their worst case execution time and low-criticality tasks can reach their deadlines, even if not enough execution time was provisioned offline. Simulations show that using the best compromise, the energy consumption can be reduced up to 17% while the percentage of deadline misses is kept under 4%.