In this work, we study two main classes of detectors for spatially multiplexed Multiple-input Multiple-output (MIMO) systems. For the first group, i.e. hard-decision detectors, we study sphere detectors, and propose novel algorithms as well as efficient architectures which make them suitable for low-complexity implementations. Furthermore, different variations of such detectors are prototyped on Xilinx FPGAs embedded on Wireless Open-access Research Platform (WARP). The second class of detectors are soft-decision detectors where, generally, soft sphere detectors are used; however, we study a new class of detectors that can serve the same purpose through a stochastic approach known as Markov Chain Monte Carlo (MCMC) technique. A general architecture with various complexity reduction techniques is proposed for this scenario, and it is shown that MCMC achieves better performance compared to sphere detector; while it requires less computation when higher order modulations are used.