The Parallel Computing Toolbox will run on any of the systems you mention. There's no specific hardware requirement beyond the base MATLAB system requirements. That said, the hardware requirements (RAM, number of cores, number of machines in the cluster, etc) are more driven by the problems that are trying to be solved than any specific hardware spec. A very generic starting point is one worker per core, and 2 GB of RAM for each worker -- but it's very easy to find reasons to need more RAM or to find out there are other bottlenecks that change the solution time. It depends on if you are running examples and learning, which you can do on pretty much any system, or if you start doing analysis with larger matrix sizes, which will start to require more RAM, and if possible, you might want to split the work up over more cores or more systems.
Another thing that PCT opens the door to is GPU computing, as well.
You might find this previous discussion interesting: http://www.mathworks.com/matlabcentral/answers/27749-cores-vs-speed-tradeoff-for-a-matlab-computer
