Experimental results of diversity estimates in a set of populations often exhibit contradictory patterns when different marker systems are used. Using simulations we identified potential causes for these discrepancies. These investigations aimed also to detect whether different sampling strategies of markers within the genome resulted in different estimates of the diversity at the whole genome level. The simulations consisted in generating a set of populations undergoing various evolutionary scenarios which differed by population size, migration rate and heterogeneity of gene flow. Population diversity was then computed for the whole genome and for subsets of loci corresponding to different marker techniques. Rank correlation between the two measures of diversity were investigated under different scenarios. We showed that the heterogeneity of genetic diversity either between loci (genomic heterogeneity, GH) or among populations (population heterogeneity, PH) varied greatly according to the evolutionary scenario considered. Furthermore, GH and PH were major determinants of the level of rank correlation between estimates of genetic diversities obtained using different kinds of markers. We found a strong positive relationship between the level of the correlation and PH, whatever the marker system. It was also shown that, when GH values were constantly low during generations, a reduced number of microsatellites was enough to predict the diversity of the whole genome, whereas when GH increased, more loci were needed to predict the diversity and amplified fragment length polymorphism markers would be more recommended in this case. Finally the results are discussed to recommend strategies for gene diversity surveys.