A comparison between the convergence of SQS OS and most other solvers in elsa can be found in the :ref:`Solvers Overview <elsa-solvers-choosing-a-solver>`.

One very tunable parameter when using ordered subset solvers is the number of subsets. Generally the more subsets used

during the optimization the faster the initial convergence speed will be. The limit to how many subsets can and should

be used for a given problem is given by the number of measurement data points.

To get a rough understanding of how the SQS OS method performs when tweaking the number of subsets we evaluated a small

3D 32 x 32 x 32 phantom reconstruction problem. The code for these experiments can be found