The emergence of GPUs has brought a revolution in the world of High Performance Computing (HPC), allowing to achieve speedups equal or greater than CPU-only computing clusters, with a significantly lower economic cost. This article presents the parallel implementation of the Pixel Purity Index (PPI) algorithm, a well-known algorithm for the analysis of remotely sensed hyperspectral data sets, targeted to a GPU cluster, and shows a comparison of execution times and speedups among different parallel implementations of the algorithm.