I've been keeping in touch with DNB developments over the years and there seems to have been a lot of conflicting research studies (wrt IQ gains, specifically the fluid intelligence component). One of the people I follow is Gwern, a member of the brain workship Google group, who's managed to gather a ton of information to help draw a "conclusion" for the lay person. See the FAQ at http://www.gwern.net/DNB%20FAQ#n-back-improves-working-memory, and a meta-analysis here http://www.gwern.net/DNB%20meta-analysis.
The section that titles "Does it really work?" suggests it does help WM, but IQ gains are more suspect.
"There are quite a few studies showing significant increases in
working memory: WM is something that can be trained. See for example
“Changes in cortical activity after training of working memory - a
single-subject analysis.” or “Increased prefrontal and parietal
activity after training of working memory”. There are a few studies
showing that DNB training enhances Gf; see the support section. There
is also a study showing that WM training (not DNB) enhances Gc65."
The meta-analysis suggests there is a medium effect, equivalent to about 2.7 IQ points.
"I meta-analyze the >19 studies which measure IQ after an n-back
intervention, confirming that there is a net gain of medium effect
"Well, the effect size from the active-control-group studies as of
today is d=0.18, and that's in standard deviations of the raw
scores, which on a normalized IQ scale with standard deviations of 15,
would then be (I think) 15 * 0.18 = +2.7 IQ points."
It is still debated how much improving working memory can improve IQ, although most accounts say the two are highly correlated. What's more certain is that playing DNB probably improves working memory. Lots of attention has been given towards reducing ADHD symptoms, presumably those lacking in working memory.
See Paul King's answer on relationship between WM and IQ.