Receivers dealing with the inter-symbol interference (ISI) caused by Faster-Than-Nyquist (FTN) signaling are often very expensive. We propose to study the simplification of a time-domain block equalizer based on Expectation Propagation (EP) by exploiting the Woodbury matrix identity. By approximating the inverse of a matrix that dominates the computational complexity to a sparse matrix that has only Lc non-zero columns, we study the performance/complexity trade-off of this solution. This allows us to get from a cubic order of complexity in block size to a quadratic complexity, allowing for instance to reach more than 83% complexity reduction for blocks of more than 1000 symbols and filters with a delay spread up to 5 symbols. The lower the FTN symbol time compression, the larger Lc must be to maintain attractive performance, which slightly decreases the complexity gain