Based on the principle of the strap-down inertial navigation system and the attitude algorithm, a novel evaluation method is proposed. It could be used to effectively diagnostic the precision of the attitude algorithm. The experimental data including Algorithm 1 (N=2, P=0, optimal two-sample), Algorithm 2 (N=2, P=1) and Algorithm 3 (N=2, P=2) is recorded. The precision of the three algorithms is evaluated by calculating the relative error. In addition, the influence of the sampling frequency on the precision is investigated. The test results show that the precision of Algorithm 2 and Algorithm 3 is approximate; as the sampling frequency is improved from 1Hz to 20Hz, the precision of the attitude algorithm can research 0.08°. When choosing the attitude algorithm of the strap-down inertial navigation system, increasing the number of the sample not always depresses the coning error, improving the sampling frequency and using the former attitude to renew the number of the sample of the period could improve the precision of the attitude algorithm, further.