In MS-OCT, where the FFT is replaced by cross-correlation, some of the preparatory steps (resampling and compensation for dispersion mismatch between the two arms of the interferometer) are no longer necessary.15–17 Thus, the MS-OCT, being based only on mathematical operations, is able, in principle, to produce better images in terms of their axial resolution and sensitivity than their FFT-based counterparts. Moreover, as MS-OCT involves mathematical operations that can be performed in parallel, it makes sense to take advantage of tools already harnessed by the OCT community to produce images fast, such as GPU (Refs. 1819202122232425–26) or FPGA (Refs. 272829–30) hardware. In terms of computation speed, GPU is an efficient computational engine for very large data sets. By contrast, the FPGA can be as efficient as the GPU for large data sets, but FPGAs are better suited for applications where the size of the data to be processed is small.31 However, on FPGAs, memory is a scarce resource as the block RAM capacity of the FPGA is generally limited. In addition, when performing cross-correlations, which are required by the MS method, data from both the current channeled spectrum and the mask must be available simultaneously as both have to be loaded from the memory in the same clock cycle, which makes the FPGA solution highly dependent on fast memory operations. This favors GPUs as the technology of choice, instead of the FPGAs, to produce TRT MS based B-scan images. Apart from the superiority of the GPU in terms of memory availability, the developers should also consider in their choice that in the foreseeable future, the multicore architecture of the CPU and the many-core architecture of the GPU are likely to merge. This trend can be seen in the introduction of computing parallel solutions used as coprocessors, such as Intel Xeon Phi and NVIDIA Tesla. Visual Studio (Windows) and GCC (Linux) both offer programming environment and language support for these architectures. A GPU combines both fine-grained (threads) and coarse-grained (blocks) parallel architectures. This feature offers an optimal mapping for the specific structure of the signal generated by the OCT systems, where data points are processed by parallel threads and spectra are processed by thread blocks. In addition, the GPU is financially and technologically a more accessible solution. Finally, there are some other immediate benefits of using GPUs over FPGAs in OCT, such as no need for extra hardware (standard PC components are sufficient), availability of a free programming environment, language support, and additional libraries (NVIDIA CUDA C, CUDA FFT).