An efficient sub-optimal algorithm, called HIS (Hyperplane Intersection and Selection) detection algorithm, is proposed to solve the problem of joint detection of K users in a MC-CDMA system. Compared to the existing solutions, the proposed algorithm has three characteristics very attractive for pratical systems. Firstly, it has nearly optimal performance. Secondly, it has a low computational complexity (O(K^2) multiplications and O(K^3) additions). Third, the algorithm has an inherent parallelism. To our knowledge, the HIS algorithm is not just an add-on to a previous existing algorithm but a rather new decoding technic based on a singular value decomposition of the channel matrix H. After giving the equation of the MC-CDMA multi-user detection problem, the HIS algorithm is described. Its performance is compared to known existing algorithms (ZF, MMSE, PIC and Sphere Decoding). For a BER as low as 10^-4 the HIS algorithm introduces only 0.2 dB degradation compared to the optimal Sphere Decoding algorithm for K = 16 users againt 3.8 dB for the PIC (with two MMSE stages) algorithm