The effects of rapid, recent uplift of the Hengduan Mountains on evolution and diversification of young floristic lineages still remain unclear. Here, we investigate diversification of three closely related Saxifraga species with a distribution restricted to the Hengduan Mountains (HM) and southern Tibet, and comment on their taxonomy based on molecular evidence. Three chloroplast DNA fragments (rbcL, trnL-F, trnS-G) and the nuclear ribosomal DNA internal transcribed spacer (ITS) were employed to study genetic structure across 104 individuals from 12 populations of Saxifraga umbellulata, S. pasumensis, and S. banmaensis. Chloroplast DNA (cpDNA) phylogenies revealed two well supported clades, corresponding to S. umbellulata and S. pasumensis plus S. banmaensis. Topology of the ITS phylogeny was largely congruent with that generated from cpDNA haplotypes, but with minor conflicts which might be caused by incomplete lineage sorting. Analyses of molecular variance of both cpDNA and ITS datasets revealed that most variation was held between S. pasumensis s.l. (with S. banmaensis) and S. umbellulata (92.31% for cpDNA; 69.78% for ITS), suggesting a high degree of genetic divergence between them. Molecular clock analysis based on ITS dataset suggested that the divergence between S. pasumensis s.l. and S. umbellulata can be dated to 8.50 Ma, probably a result of vicariant allopatric diversification associated with the uplift events of the HM. Vicariance associated with HM uplifts may also have been responsible for infraspecific differentiation in S. pasumensis. In contrast, infraspecific differentiation in S. umbellulata was most likely triggered by Quaternary glaciations. The much lower levels of gene diversity within populations of S. pasumensis compared with S. umbellulata could have resulted from both range contractions and human collection on account of its putative medicinal properties. Combining evidence from morphology, geographical distributions and molecular phylogenetic data, we recommend that S. banmaensis should be treated as a synonym of S. pasumensis which in turn, and based on the same sources of evidence, should be treated as a separate species rather than as a variety of S. umbellulata.