One may find on the web a lot of different computational packages to perform "ab-initio" calculations of electron structure of the solids. Usually, the documentation is not quite transparent about the disadvantages of the methods used. It takes some time to find out whether given package is suitable for you needs. And even if it is according to the documentation, it may turn out that it is not in reality.
Is there a good comparison of different packages? Under "comparison" I mean not usability, etc. but comparison from the researcher's point of view.
One aspect is the purpose of the package. I.e., some packages (like GAUSSIAN) are more chemistry-oriented, some focus on structural properties, and neither may be used if you want to calculate e.g. exchange interaction between the excitons in semiconductors. However, you may spend days realizing this fact from the documentation because authors did not know that such a problem even exists.
Another aspect is methods which are implemented. The documentation is usually quite sparse on this subject also due to the fact that authors are not well aware of other methods which exist around. A good example is spin-related phenomena. Many packages introduce some limited tricks to handle these problems because they never thought that there are problems where these tricks do not work.
Few examples: abinit, gaussian, quantum espresso, siesta, VASP, WIEN2k, Octopus
This post has been migrated from (A51.SE)