Let y = x - 8 and then cube both sides (this isn't strictly necessary, but makes some of the intermediate working easier because symmetry). The equation becomes (note that I will use cbrt for cube roots throughout):
cbrt(y + 8) + cbrt(y - 8) = cbrt(y)
(y + 8) + (y - 8) + 3 [cbrt((y + 8)(y - 8)^2) + cbrt((y + 8)^2(y - 8))] = y
or, simplifying,
y + 3 cbrt(y^2 - 64) [cbrt(y + 8) + cbrt(y - 8)] = 0
by stealing the starting equation, we can reduce this to just
y + 3 cbrt(y^2 - 64)*cbrt(y) = 0
and then rearrange (taking the cube-root stuff to the right-hand side) and cube again:
y^3 = -27 y (y^2 - 64) .
The first solution is just y = 0 (or x = 8, recovering Hymie's solution). The second possibility is that
y^2 = -27 (y^2 - 64)
=> y^2 = 27*16/7 = (3/7)*144
=> y = ±12√(3/7),
from which x = 8 ±12√(3/7) forms a second pair of solutions.
We should formally check that these also work, since it's often possible to introduce "false" solutions by "undoing" roots or squaring square roots. However, we're saved from having to do this here because the cube root is a bijective function over real numbers (ie, for every real-numbered input there's a unique real-numbered output).
Still, if you check, then you find that the equation does work for these other solutions, with left and right-hand sides being about ±1.9879 for x = 8 ±12√(3/7) respectively.
In the end, if you didn't bother to do the y = x - 8 thing, then the only extra complication would be having to solve the quadratic equation
(x - 8)^2 = -27x(x - 16),
which rearranges to
7x^2 - 112 x + 16 = 0,
with the same solutions as before. But it's much, much easier to go wrong doing it this way, by hand at least, so rearranging to start with makes things a helluva lot easier.