We present a Monte Carlo (MC) radiative transfer code for complex three dimensional dust distributions and include transiently heated PAH. The correctness of the code is confirmed by comparison with benchmark results. The method makes use of the parallelization capabilities of modern vectorized computing units like graphic cards. The computational speed grows linearly with the number of graphical processing units (GPU). On a conventional desktop PC, our code is up to a factor 100 faster when compared to other MC algorithms. As an example, we compute the dust emission of proto-planetary disks. We simulate how a mid-IR instrument mounted at a future 42 m ELT will detect such disks. Two cases are distinguished: a homogeneous disk and a disk with an outward migrating planet, producing a gap and a spiral density wave. We find that the resulting mid-IR spectra of both disks are almost identical. However, they can be distinguished at those wavelengths by coronographic, dual-band imaging. Finally, the emission of PAHs exposed to different radiation fields is computed. We demonstrate that PAH emission depends not only on the strength but also strongly on the hardness of the radiation, a fact which has often been neglected in previous models. We find that hard photons (>20  eV) easily dissociate all PAHs in the disks of T Tauri stars. To explain the low, but not negligible detection rate (<10%) of PAHs in T Tau disks, we suggest that turbulent motions act as a possible path for PAH survival.