Aims: The main objective of this study is to quantify the genetic variability and relationship among the quantitative traits of the torch ginger germplasm collected across Peninsular Malaysia. Study design: Initially, the experimental design was randomized complete block design (RCBD) but there were unequal number of replications due to limited planting materials. Place and Duration of Study: The study was conducted at the Malaysian Agricultural Research and Development Institute (MARDI) Research Station, Jerangau, Terengganu, Malaysia on latitude 04°57.704’N and longitude 103°11.007’E. Data on agro-morphological characterization were collected for three years from June 2010 until May 2013. Methodology: A total of 57 torch ginger accessions were collected from various wild/ cultivated sources from seven states of Peninsular Malaysia namely Terengganu (19), Perak (15), Kedah (7), Johor (6), Pahang (5), Kelantan (3) and Melaka (2). The collection was planted in 2009 and maintained as living collection. At three months old torch ginger plantlets were transplanted into 20 cm × 20 cm × 20 cm holes spaced at 100 cm diameter of culvert. The spacing of culvert between rows and within rows were 2 m respectively. Initially, the experimental design was randomized complete block design (RCBD) but there were unequal number of replications due to limited planting materials. Data on agro-morphological characterization were collected for three years from June 2010 until May 2013. All the 57 accessions were characterized according to descriptors list of genera under Zingiberaceae family with some modifications. A total of 6 qualitative and 16 quantitative descriptors were used. The mean values of each quantitative trait data were computed and subjected to statistical analysis to assess the amount of genetic variation using analysis of variance (ANOVA). The ANOVA was using PROC GLM of SAS 9.4 software (SAS Institute Inc., Cary, NC, USA). Genotypic coefficients of variation and error variance were estimated using PROC VARCOMP method Type I of SAS 9.4 software. Phenotypic coefficients of variation, heritability and genetic advance were calculated. Pearson correlation coefficient was determined using SAS 9.4 for comparing the relationship among the different traits. The morphological traits were analyzed by numerical taxonomic methods via cluster analysis and Principal Component Analysis (PCA) which was done using NTSYS-PC Version 2.1. PCA based on 16 quantitative and six qualitative traits was performed to determine the genetic variation in the torch ginger accessions. The cluster and PCA were performed to reveal clustering and grouping pattern of torch ginger accessions. Euclidean distance coefficients were estimated for all pairs of accessions, which was then used to present cluster analysis. Results: Analysis of variance indicated that highly significant variation exists among the 57 accessions for most of the traits. The 57 accessions of torch ginger showed high variability for both qualitative and quantitative traits. Among the quantitative traits, the highest CV was observed from number of spikes with the value of 44.25%. The genetic coefficient of variation for 16 morphological traits ranged from 9.76 to 45.86%. Broad sense heritability estimates varied from low to high. Maximum estimates of broad sense heritability were recorded in bract length (80.16%), bud peduncle length (78.74%), number of leaves (75.66%), number of stems per clump (67.99%), inflorescence peduncle length (67.60%) and bud width (65.11%). The genetic advance (GA) in percent of mean was recorded from 11.82 to 67.97%. Besides that, positive correlation coefficient was obtained between some traits. Generally, the genotypes groups were related with morphological characteristics among the accessions. Conclusion: Considering the distribution pattern, vegetative and yield performance of accessions, it is suggested that accessions from cluster VII (KAN022, KAN047 and KAN048) should be selected for future breeding program. These accessions possessed the highest peduncle length that could influence the yield in terms of bud size. Furthermore, the classification and divergence between torch ginger accessions analyzed in this study may assist in conserving plant materials both in-situ and ex-situ.