This paper studies the multi-carrier two-user Gaussian Interference Channel (G-IFC) with an adaptive strategy depending on the channel gains. The decoding strategy at each receiver is either Successive Interference Cancellation or treating interference as noise, depending on whether interference is strong or weak. The transmit strategy may also rely on FDMA when required. Then, an iterative power allocation algorithm is proposed in order to maximize the sum rate when considering all carriers jointly. By splitting power allocation at both transmitters, the initial optimization problem can be turned into a series of convex optimization problems. The proposed algorithm converges very rapidly and provides a sum rate increase compared to FDMA and to a strategy where interference is always considered as noise.