One major challenge in the bioconversion of lignocelluloses into ethanol is to develop Saccharomyces cerevisiae strains that can utilize all available sugars in biomass hydrolysates, especially the D-xylose and l-arabinose that cannot be fermented by the S. cerevisiae strain naturally. Here, we integrated an l-arabinose utilization cassette (AUC) into the genome of an efficient D-xylose fermenting industrial diploid S. cerevisiae strain CIBTS0735 to make strain CIBTS1972. After evolving on arabinose, CIBTS1974 with excellent fermentation capacity was obtained. A comparison between genome sequences of strains CIBTS1974 and CIBTS1972 revealed that the copy number of the AUC had increased from 1 to 12. We then constructed the AUC null-mutant CIBTS1975 and gradually rescued the l-arabinose utilization defect by integrating AUC iteratively. On the other hand, the parental strain CIBTS0735 was able to acquire the same performance as CIBTS1974 by the direct introduction of 12 copies of the AUC; the performance was further improved by adding two more copies. Besides, we found that not the two transporters present in the AUC were both needed during l-arabinose utilization, GAL2 was necessary and STP2 was not essential. We have described for the first time that a high copy number of AUC is sufficient for the strain to metabolize l-arabinose efficiently independent of evolution.