Modern System-on-Chip (SOC) architectures offer much for a relatively small price, but often industrial machine builders only use a fraction of the functionality. Their main interest is the performance boost by using multiple cores. For safety devices, the on-chip redundancy is beneficially to achieve higher reliability, but since most platforms are homogenous, there is a need to get systematic and common cause failures under control. Graphics Processing Units (GPU`s) are more and more available in modern SOCs, but their utilization in industrial products is still marginal. This paper proposes a concept how GPUs can improve the overall safety integrity with additional diversity in hardware and software.