We still have trouble with unanswered questions, as has been discussed previously. At the moment, the total is 3257 unanswered out of 27476 total questions, or nearly 12%. Regardless of what we do, this situation does not seem to have changed very much with time: always around 10% are unanswered.
I think that part of the problem is that some questions are not obviously unclear or irrelevant to our topic, but nonetheless embody subtle misconceptions that make it impossible to answer productively. A direct answer to this type of question requires acceptance of those misconceptions at face value, while an attempt to guide the querent in a more productive direction would distinctly not be an answer, and in fact would render the question obsolete.
Let us take this one as an example. Here, OP asks for something that certainly has a direct answer, although not in a very compelling or well-defined way (for example, benchmark code should have been given, otherwise the exercise is arbitrary and it will not be meaningful to compare the results). But whatever answers one may give certainly will not be of any use in actually doing GPU-based computations in Mathematica, because the idea behind the question is overly simplistic.
Such questions arise, I think, because people will often ask a question based on an early understanding of their problem. In this sense they are highly localized, but because the OP's assumptions are normally implicit, it is difficult to express that clearly enough in relation to the question content to argue for closure. If the misconceptions related to Mathematica itself, we would have no hesitation or difficulty in closing the question as a simple mistake. The situation is objectively no different if the misconception lies in another field, but since such a question might be perfectly well defined in its relationship to Mathematica, to express this may seem improper.
I have decided to try an experiment and used an unusual argument for closure: that to answer the question would not be productive. Obviously, this opinion is highly subjective, and I would not advocate its unrestrained use. But for long-standing unanswered questions that one strongly feels are ill-posed or misguided, I cannot think of a better close reason, even though it is clearly pointless to have such questions remain open and unanswered forever.
I stop short of a call to action based on this idea, but I would certainly like to hear others' opinions about it. Then, if we decide to use it as an actual close reason, we will be able to link back to this thread to help explain the rationale behind it.