Massive MIMO systems employing hundreds of antennas at the base station (BS) are considered a breakthrough technology to provide users with high data rates. However, large number of antennas demands memories of large size which are prone to faults due to current aggressive technology downscaling. This paper introduces a novel nonlinear minimum mean square error (NMMSE) based detection algorithm that takes memory errors into account. The proposed detection method is able to handle multiple memory errors with low computational overhead. Simulation reports that the proposed solution significantly reduces the impact of multiple memory errors on the bit error rate (BER).