This paper presents a trade-off analysis associated with different matrix decomposition algorithms in the application of MIMO precoding. A scenario of 5G New Radio (NR) Multi-User MIMO (MU-MIMO) downlink involving one base station (BS) equipped with 64 antennas which serves multiple users simultaneously is studied. The channel model follows the 3GPP TDL-C with various levels of antenna correlation. Six different matrix decomposition algorithms are studied, namely QR Modified Gram-Schmidt (QR-MGS), QR householder, Cholesky, LDL, SVD Bidiagonalization, and SVD. An evaluation of the computational complexity of the algorithms is presented and the system performance compared. These analyses allow us to make an informed trade-off between performance and hardware complexity.