This paper describes a novel low complexity scalable multiple-input multiple-output (MIMO) detector that does not require preprocessing and the optimal squared l2-norm computations to achieve good bit error (BER) performance. Unlike existing detectors such as Flexsphere that use preprocessing before MIMO detection to improve performance, the proposed detector instead performs multiple search passes, where each search pass detects the transmit stream with a different permuted detection order. In addition, to reduce the number of multipliers required in the design, we use l1-norm in place of the optimal squared l2-norm. To ameliorate the BER performance loss due to l1- norm, we propose squaring then scaling the l1-norm. By changing the number of parallel search passes and using norm scaling, we show that this design achieves comparable performance to Flexsphere with reduced resource requirement or achieves BER performance close to exhaustive search with increased resource requirement.