This paper considers the uncoded multiple-input multiple-output (MIMO) system with linear minimum mean square error (MMSE) detection under ideal fast fading. The distribution of SINR at the output of the MMSE detection is derived for a small number of transmit and receive antennas. We present new approximation for the Gaussian Q-function driven by numerical simulation. Based on the SINR distribution and new approximation for Q-function, we analyze the performance of linear MMSE detection under ideal fast fading environment. By comparing the analytical results and Monte Carlo simulated results, we validate the analytical results.