It's hard to know how to evaluate much of this in a global way. There are many ways of assessing "goodness". Here, however, are a couple I have found.
Note I am a statistician, data scientist, and quantitative engineer. I work a lot with numbers and with quantitative and highly structured ways of dealing with text. My world is basically Python 3 or R, and R heavily dominates. I can run either on Ubuntu or on Windows, and I use, at various times, 7 Home Edition, 7 Enterprise ("Pro"), and 10.
Python is Python, but its problem is that numpy and scipy do not have the numerical rigor instilled in them seen in a MATLAB or in R. Python has a lot of packages and I will sometimes use it, but primarily to prepare datasets for R.
R is my main, so it's my world.
The principal problems with R on Ubuntu vs Windows are:
(a) On Ubuntu, new packages are expected to be recompiled. Accordingly the outcome of an install is not only dependent upon the new functions being downloaded, but, during compilation, the library base and the compiler versions available. Sometimes these fail. On Windows, the default means of distribution is to distribute binaries. This freezes all interrelationships among packages into a solid, consistent, and mostly often successful mixture.
(b) Memory management on Ubuntu -- even with multiple cores -- can be complicated. If I have a calculation which demands 100 Gb of working space, a choice of a fixed memory.limit specification can be lethal. Moreover, much of the focus upon software like Python is to be able to script calculations. This isn't always the best way to do things.
