The cassava core collection was selected to represent, with minimum repetitiveness, the potential genetic diversity of the crop. The core (630 accessions) was chosen from the base collection (over 5500 accessions) on the basis of diversity of origin (country and geographic), morphology, isozyme patterns and specific agronomic criteria. To asses the genetic diversity of the core, 521 accessions were typed with four microsatellite loci. Allele diversity and frequency, and size variance of dinucleotide repeats (Rst statistic) were estimated. Microsatellite allele numbers and frequencies varied among countries: Colombia and Brazil had the largest number of different alleles across all loci. Mexico also had a high number, ranking fifth after Peru, Costa Rica and Venezuela (which tied). Unique alleles were present in accessions from Brazil, Colombia, Guatemala, Venezuela and Paraguay. A small number (1.34%) of potential duplicates were identified through isozyme and AFLP profiles. Thus, the present results indicated that traditional markers have been highly effective at selecting unique genotypes for the core. Future selections of cassava germplasm sets can be aided by DNA-based markers to ensure genetically representative, non-redundant samples.