Two methods which allow to solve non-symmetric sparsed systems of linear equations (SLEs) - Biconjugative Gradients (BiCG) and General Minimal Residuals (GMRES) methods - are described in this paper. Analysis of requirements to memory and computational speed is given; some results of application to finite- differences and finiteelements SLEs are shown. Some features of these methods connected with ABS-class presented by Abaffi and Spedicato are also discussed