Some context: Several frequently cited studies on working memory training using dual n-back, most famously Jaeggi et al. 2008 strongly indicated that WMT could reliably produce lasting benefits to fluid intelligence. These studies obviously provide great material for marketing cognitive training software. See for instance this page by Lumosity:

In 2008, Dr. Susanne Jaeggi, Dr. Martin Buschkuehl and colleagues at the University of Michigan showed that cognitive training with a task called Dual N-Back enhanced fluid intelligence – the ability to creatively solve new problems, and a critical component of IQ. This study involved healthy young adults, mostly university students. After as little as eight hours of training, young adults who trained saw significant gains in fluid intelligence and working memory. We have worked with the Michigan group to include a version of their training program on Lumosity. In addition, we have created a game-like version of their task called Memory Lane.

However, later studies seem to have found little evidence in support of these claims. For example, the 2012 meta-analysis Is Working Memory Training Effective? concludes:

The absence of transfer to tasks that are unlike the training tasks shows that there is no evidence these programs are suitable as methods of treatment for children with developmental cognitive disorders or as ways of effecting general improvements in adults’ or children’s cognitive skills or scholastic attainments.

To me, it seems that the case for dual-n back exercises actually yielding transferable improvements to intelligence and memory is much weaker after some years of scrutiny. The linked dual-n back game seems like an excellent alternative to existing n-back software. I'm not sure it's worth the time and effort.