We derived the LTE abundance of silicon from 65 lines of Si I using one-dimensional semi-empirical models of the solar atmosphere HOLMUL, MACKKL and VAL,C. The sample of the lines is considerably larger than previously exploited. We confirmed a reliability of the solar oscillator strength scale of Gurtovenko and Kostik for Si I lines which is obtained from fitting to the observed equivalent widths. We show that this scale is displaced from the experimental scales derived by Becker et al. and Garz by +0.073 dex and –0.026 dex, respectively. The differences between solar and experimental oscillator strengths of individual Si I lines do not depend on their lower excitation potentials, wavelengths and equivalent widths. Such a shift of the solar scale results mainly from ignoring NLTE and granulation inhomogeneity effects, from the choice of the one-dimensional atmospheric model, from uncertainties of the broadening damping constant, microturbulence and observed equivalent widths. We investigated the effect of changes in the various input parameters on the fitted LTE-abundance. It is shown that both the experimental oscillator strength scale of Becker et al. and solar one shifted by +0.073 dex produce almost the same silicon abundance. The total rms error of the abundance which is caused by errors in equivalent widths and microturbulence velocity is 0.02 dex. Employing semi-classical theory of Anstee, Barklem and O’Mara for determination of van der Waals broadening produces a trend with line strength in the abundances fitted from Si I lines. It is cancelled when using Unsöld equation multiplied by an enhancement factor of E = 1.5. The average value of the abundance difference caused by different damping treatment does not exceed 0.03 dex. With HOLMUL model, E = 1.5, and solar oscillator strengths attached to the experimental scale of Becker et al., the derived silicon LTE-abundance equals 7.594±0.015, while with VAL,C model it is 7.623±0.021.