This is generally a good starting point, but it's not something that is a rule that is going to be true for every problem being solved. Some problems will benefit from more cores. Others will benefit from increased memory, network bandwidth, faster disks, or some combination of the above. Some may even do better with less cores!
It's also important to know if the cores are indeed real cores, or the result of hyper threading. You may see a performance benefit by turning off hyper threading in some cases, or only using the number of real cores that you have.
I'd recommend that you do some runs of your problem with different numbers of workers and graph the performance. It's likely that you will be able to see a point at which additional workers don't provide any additional benefit.
If you take a look at the "Benchmarks and Performance" section of the PCT Demos page, it might give you some good starting points for analyzing your problem.
http://www.mathworks.com/products/parallel-computing/demos.html
