The model of homogeneous spheroids were chosen to provide the detailed comparison of two popular solutions of the light scattering problem: T-Matrix Method and Discrete Dipole Approximation. The exact solution by the Separation of Variables Method were used as a standard giving the most accurate results. We have computed the scattering cross-sections of prolate and oblate spheroids with the refractive index m = 1.3 and 2.5 at fixed orientation in a wide range of the aspect ratios and sizes. We found that: i) the coincidence of the T-Matrix Method and the Separation of Variables Method is very good (> 6 - 10 digits) up to some boundary particle size; for larger particles the precision of T-Matrix results sharply drops; ii) the Discrete Dipole Approximation code gives the satisfactory results (the deviations from other methods less 5 - 10 %) for large values of size and aspect ratio even if the number of dipoles is 1, 000 - 1, 500; the accuracy less than 1 % may be obtained if the number of dipoles exceed 10, 000 -50, 000; iii) the accuracy of the methods decreases with the growth of the parameter τ = m · (2πr_v/λ) · (a/b), where r_v is the radius of equivolume sphere, λ the wavelength of incident radiation, a/b the aspect ratio. If a/b ≤ 4, the coincidence of the results with those of the Separation of Variables Method is within 1 - 3 % for τ ≈ 8 - 16 (Discrete Dipole Approximation) and τ ≈ 50 - 65 (T-Matrix Method). For the particles with a/b ≥ 10, the Separation of Variables Method is preferable, if 2πr_v/λ ≥ 2 - 3.