This paper presents a comparison and analysis of parallel sequential and parallel implementations for algebraic reconstruction algorithms. These algorithms are the fundamental building block used for reconstructing water vapor images from Global Navigation Satellite System (GNSS) observations. In order to obtain real-time reconstructions from this data high performance software needed to be developed, the various libraries and parallelization techniques available makes it important to study the underlying performance of the algorithms under competing conditions. In this paper we compare multithreading CPU parallelization’s techniques with the performance that can be obtained using GPU parallelization techniques. The results obtained indicate that Algebraic Reconstruction Techniques (ART) algorithm approaches like Kaczmarz do not obtain much benefit from the GPU parallelization while Simultaneous Iterative Techniques (SIRT) algorithm approaches like Landweber have a large benefit from being parallelized in CPU and GPU.