I feel nowadays to meaningfully talk about performance of N-queens problem N should start with 18 instead of 8.. Here is a C++ solution with 40 lines of code and can solve 18-queens in less than a minute on a laptop:
That'll do ~N=1e6 in a second (I'm sure there are lots of optimisations that could be done in it though), or solutions for N=4 to N=1000 in less than 10.
The more complex problem is finding out how many solutions there are in total.
its in ~210 lines of C++:
https://bitbucket.org/rfc/dopt_nqueen/src/3e812164d97a8e91d4...
uses random local search approach, as described in russell and norvig (http://aima.cs.berkeley.edu/).
wrote this code during https://www.coursera.org/course/optimization
That'll do ~N=1e6 in a second (I'm sure there are lots of optimisations that could be done in it though), or solutions for N=4 to N=1000 in less than 10.
The more complex problem is finding out how many solutions there are in total.