Large extent of Copy Number Variants (CNV), i.e changes in the copy number of genes between individuals, have been recently highlighted in maize using Comparative Genomic Hybridization array or massive parallel sequencing. However, their contribution to genetic diversity and to traits variation remains mostly unknown since these technologies are very expensive and labor demanding. To address these issues, we developed an original approach based on Affymetrix Axiom technology, to genotype an extreme form of CNV called Present Absent Variant (PAV) in maize. PAV was defined as DNA sequence >1kbp that is present in some individuals but absent from others. Using our high throughtput genotyping array, we genotyped 60 026 PAV on 356 inbred lines from an association panel representing worldwide maize genetic diversity. This panel has been previously genotyped using 50k SNP Illumina Infinium array and phenotyped for 29 agronomic traits related to yield, phenology and plant architecture. We analyzed and compared how PAV and SNP polymorphisms were globally structured in these panel by analyzing relatedness and genetic structuration. We observed that genetic structuration and relatedness obtained using PAV were globally similar to those obtained with SNP. We analyzed extent of linkage disequilibrium (LD) between SNP and PAV. We observed that LD were less extended between PAV and SNP than between SNP suggesting that the effect of these polymorphisms on traits could be not easily captured by linkage disequilibrium with SNP from 50K array. We performed a genome wide association study on 29 agronomic traits and identified several PAV significantly associated with different agronomic traits.