The fast development of array technology has raised the density of oligonucleotide SNP arrays from 10K and 50K to 100K and 500K. However, methods for SNP genotyping have not been developed as fast. Most methods are based on sample-dependent multi-array training and may not be suitable for cross-laboratory studies and small sample studies, few use full information of array technology efficiently. It has been suggested that mismatch probes are of no use in genotyping and thus should be removed, and that multiple probes are redundant and should be reduced. We study the effect of reduced probes on genotype calling and conclude that reduction of probes in SNP interrogation substantially increases genotyping error. Furthermore, elimination of mismatch probes increases error rate by 50 folds. Thus caution should be used in SNP array design. Array design with fewer probes for more SNPs in single array may sacrifice genotyping quality.