In this paper we get a power estimate from above of the probability that
Buffon's needle will land within distance 3^{-n} of Sierpinski's gasket of
Hausdorff dimension 1. In comparison with the case of 1/4 corner Cantor set
considered in Nazarov, Peres, and the second author: we still need the
technique of arXiv:0801.2942 for splitting the directions to good and bad ones,
but the case of Sierpinski gasket is considerably more generic and lacks
symmetry, resulting in a need for much more careful estimates of zeros of the
Fourier transform of Cantor measure.