This contribution proposes a method of MRL simplification which saves on the line buffer for above-left reference samples. MRL intra prediction uses 2 additional reference lines for intra prediction, which results in larger line/columns buffer for reference samples. The proposed method uses one sample instead of eleven samples in above left reference samples. It is reported the proposed method saves 18.6 % line buffer for reference samples with 0.03 % / 0.03 % / 0.00 % coding loss in IO / RA / LDB configuration respectively.