The point here is not "Write as many lines of code as you can". I want the candidate to plan - it's a crucial part of the job. FizzBuzz, sorts and the like don't require much planning. Minesweeper does.
Then ask for that, and assess the plan, otherwise you're expecting Carnac the Magnificent: http://raganwald.com/2015/05/08/carnac-the-magnificent.html
I did plan...but if you were watching me you would not have seen me planning.
I decided right away that my display was going to be a table with one table cell per Minesweeper cell, with the table cell's ID of the form "X_Y" where X and Y are the coordinates of the corresponding Minesweeper cell. I did the rest of my planning while getting that HTML written (first in vim, and then when that was going too slow even with macro assistance by writing a little Perl script to generate the HTML).
I was not able to do it in an hour. At the end of one hour, I had everything for a rudimentary Minesweeper coded, but had a bug in the code that reveals the cells around a clicked empty cell, and it took another 20 minutes to find and fix that.
Here is what I came up with: https://github.com/tzs/interview_minesweeper
The initial commit was made at the one hour point, so that would have been what I would have turned in during an interview.
I'd rather work with the person who would approach that by taking 3 hours to plan, 3 hours to code, and 2 hours to bug fix, rather than the "Ready, Fire, Aim" person who took 1 hour to code, 4 hours to fix what he did, 8 more hours to find his design was totally wrong and re-do major portions of it, 8 more fixing the bugs in that version........
Unless you plan to filter out the people who just jumped into the code, in which the exercise is just a trick question.