|  | I found one reference, which has only this single example: 
Find a set of 11 integral squares that can be assembled into a 13x13 square.
----------------------------------------------------------------------------
Squares of primes are generally difficult to break up efficiently.
Squares with composite sides can usually be profitably reduced to a set of
squares whose sides are divisors plus the dissection of one of those 
squares; i.e. 100^2 = 90^2 + 18*10^2 + (number of squares to make up a 
10x10).
If the side N is a power of 2, the square can be broken up into 3log[2]N+1
smaller squares.
Lynn 
 | 
|  |     To elaborate on what Lynn said in .1, you can set some upper limits
    on the minumum that you are searching for as follows:
    
    If [ N = (2^T) * K ] and (2^T) is the highest power of 2 that divides
    N, : Then we can set some upper limit U(N):
    
(1)	U(N) = 3T + 1			Where N = 2^T
    
(2)	U(N) = 3T + N			Where T > 0
    
(3)	U(N) = N + 3			Where T = 0
    
    These are probably far from optimal, except for (1), but they provide
    a good starting point.
    
    --  Barry
 |