You are free to use any (reasonable) intervals. Powers of 10 would be sufficient for this assignment, but could skip/ hide some effects in practice.
Generally speaking, using powers of 2 usually gives a good balance between computational effort (10 samples for each three orders of magnitude), and accuracy.
You could also do a coarse sampling first, and then add additional points in regions of interest (high gradients in the plot, around cache limits, ...). But this is also not required in this assignment.