was actually XP- and I abandoned that, ultimately, for Ubuntu, because I found myself spending way too much time on system maintenance- registry issues, cleaning up malware, rebuilding the system because it just became unstable for no obvious reason. Upgrading a Windows system (beginning with Win 3.1) generally resulted in breaking critical apps, and rendering ancillary hardware non functional.
I stopped upgrading my Ubuntu at 10.04, because of personal insecurity over the Unity/GNOME 3 trends- going directions I have no desire to go. But, back when I would upgrade the system, it was generally painless and straight forward, as installing a Linux distro on a new system (no need to seek out obscure drivers, etc. to get my hardware to work- everything just works nicely) is straight-forward and simple. Most importantly, I have never had a problem with legacy software (or hardware) working with a Linux update.
As to quality of software, within the scientific community, the best, BY FAR, solutions are generally Open Source and developed on *.nix platforms. Such outfits as NIST, Argonne National Laboratories, EDF of France, as well as many universities throughout the world, release scientific analysis packages that outperform, in terms of both capabilities and speed, most any commercial package I have explored. For you CAD jockeys out there, brlCAD, released as an Open Source package by the US Military Ballistics Research Laboratory, was doing things 20 years ago that commercial 3D modeling systems are only now starting to adopt- including very sophisticated ray tracing, etc. What a lot of Open Source software lacks is the fancy eye candy of the GUI, which actually interferes with performance...