FarmCPU iteraltively uses fixed and random effect models for powerful and efficient genome-wide association studies (GWAS). For a phenoype that simulated with 50% heritability and controlled by 500 eauqlly contributted quantitative trait nucleoitides (QTNs), FarmCPU could detect 50 more QTNs compared with mixed linear model (MLM) under 10% false discovery rate (FDR). A dataset with half million individuals and half million markers can be analyzed by FarmCPU within three days.