Several molecular marker systems have been developed for assessing genetic diversity in crop germplasm collections. A trade-off often exists between the number of loci that can feasibly be sampled by a marker system and the amount of information provided by each locus. We compared the usefulness of two marker systems for revealing genetic diversity and population structure in cassava (Manihot esculenta Crantz): simple sequence repeats (SSRs) and diversity array technology (DArT) markers. DArTs survey many more loci per reaction than do SSRs; however, as bi-allelic, dominant markers, DArTs provide less polymorphism information per locus. Genetic differentiation was assessed in a randomly selected set of 436 cassava accessions, consisting of 155 African and 281 Latin American accessions. A genome-wide set of 36 SSR markers and a DArT array of approximately 1000 polymorphic clones were used to assess genetic diversity and differentiation. Cluster analyses were performed using principal coordinate analysis (PCoA). Results were compared with a priori expectations of genetic differentiation based on previous genetic analyses. Analyses of the two datasets generated broadly similar clustering patterns. However, SSRs revealed greater differentiation than DArTs, and more effectively recovered patterns of genetic differentiation observed in previous analyses (differentiation between Latin American and African accessions, and some geographical differentiation within each of these groups). These results suggest that SSR markers, while low throughput in comparison with DArTs, are relatively better at detecting genetic differentiation in cassava germplasm collections. Nonetheless, DArTs will likely prove useful in ‘orphan crop’ species, where alternative molecular markers have not been developed.