In the paper the problem of constructing efficient DFT algorithms for vector sizes N=2^s 5^t is analysed. Several structures are compared: prime factor FFT, radix-10 and split-radix-10 FFTs, and nesting algorithms, including algorithms using `small' 25-point DFT module. It is shown that fairly simple algorithms are not less efficient than the radix-2 FFT for comparable vector sizes, and that their performance improve with introduction of more sophisticated ideas.