We can only speculate about the various automatic algorithms that are available to generate names for chemical structures. Anyway, all known programs have limitations or shortcomings. In particular, the algorithms may not include all rules or they may use obsolete rules. Note that the IUPAC recommendations have changed several times and that the current (2013) recommendations have more than 1500 pages.
In my experience and also in this case, ACD/ChemSketch is more reliable than MarvinSketch. However, such programs are typically updated frequently, and some issues might be solved in a new version.
The compounds that are given in the question are not trivial cases. In order to find the preferred IUPAC names, we have to dig a bit deeper and consider the rules given in Section P-45 of the current recommendations. These rules are applied if two or more names may result according to the basic nomenclature rules, because of different substitution patterns or multiple occurrences of the same senior parent structure. Apparently, these rules are taken into account in the present version of ACD/ChemSketch but not in the present version of MarvinSketch.
Thus, ACD/ChemSketch correctly gives the preferred IUPAC name for the first compound as 4-(2-bromobutyl)-6-methyloctan-3-ol.
The name ‘6-bromo-4-(2-methylbutyl)octan-3-ol’, which was proposed by MarvinSketch, is not wrong, it is just not the preferred name. Remarkably, this name was in accordance with obsolete IUPAC recommendations from 1979 (retained in 1993). Therefore, some homework exercises might still expect this answer. However, this does not necessarily mean that MarvinSketch is using old IUPAC recommendations from 1979 or 1993; it could be a mere coincidence as the second example below shows.
ACD/ChemSketch also correctly gives the preferred IUPAC name for the second compound as 4-(2-bromobutyl)-6,6-dimethyloctan-3-ol.
The name ‘6-bromo-4-(2,2-dimethylbutyl)octan-3-ol’, which was proposed by MarvinSketch, is not in accordance with IUPAC recommendations (neither 1979 nor 1993 nor 2013). Apparently, the relevant (past or present) tie-breaking rules for such cases are not included at all in the present version of MarvinSketch.