This work presents the parallel performance of the SOD2D code (Spectral high-Order coDe 2 solve partial Differential equations) [1], a new Continuous Galerkin High-Order Spectral Element Method (CG-SEM) code designed to solve simulations of both turbulent compressible and incompressible flows. The current trend in new High-Performance Computing (HPC) platforms is to achieve higher computing performances by installing increasingly more modules based on General-Purpose Graphical Processing Units (GPGPUs) [2]. Hence, the new scientific computing codes must be programmed in such a way that they can benefit from the fine-grained parallelism and large data throughput of these devices. SOD2D has been coded following this new paradigm, and the present work aims to assess and analyse the parallel performance and scalability of SOD2D code, focusing on when running on GPUs. Preliminary results show an almost ideal parallel efficiency in the weak speedup for all the computed GPU load ratios, in both compressible and incompressible solvers. Regarding the strong speedup analysis, the results are also very good, although these are strongly affected by the GPU load. The strong speedup is highly degraded in cases where the load per GPU is small and the graphical unit is not performing at its optimal point. Furthermore, the performance of SOD2D when using MPI versus NCCL (NVIDIA Collective Communications Library) implementations is analysed and compared, showing the potential and limitations of each one.