In order to study the influence between the number of markers applied or corresponding polymorphic alleles investigated and the reliability of evaluation the genetic diversity of G. hirsutum L. races using SSR markers, significance tests were done base on coefficient of variation(CV) about genetic similarity coefficient, the correlation among cophenetic matrices were also examined using Mantel correlation test. A total of 90 pairs of SSR primers were used to amplify the 42 accessions of germplasm. 440 polymorphic bands were obtained, which accounting for 88.5% of the total 530 bands. For 90 primers, the polymorphism information content (PIC) was from 0.046 to 0.888, with the average of 0.649. The Shannon-Weaver diversity index (H′) ranged from 0.113 to 2.289 and averaged at 1.248. Significance test showed that relatively accurate results could be got by 25 primers or 150 polymorphic alleles when the primers were sorted based on PIC in descending order and by at least 50 primers or 200 polymorphic alleles in ascending sort order. The correlation among cophenetic matrices was also examined using Mantel correlation test. The objectively genetic relationships could be reflected with 20 primers in descending sort order or 50 primers in ascending sort order or all 150 alleles. In addition to this, the primer offered more information which has provided more alleles when less primer is used relatively, but the information of each primer provided was becoming not obvious with the adding of primers. When using less allele, the quantity of primer is more important, and the information is more important than quantity of primer as the alleles increase. In general, it’s necessary to use 30 pairs of primers and amplify 150 alleles, the objectively genetic relationships of 42 accessions of G. hirsutum L. races could be reflected, using more than 50 is better.