Defining plausible causal variation by fine-mapping can be more effective in multi-ethnic genetic studies, particularly in regions such as the MHC with highly population-specific structure. To enable such studies, we constructed a large (N=21,546) high resolution HLA reference panel spanning five global populations based on whole-genome sequences. We jointly analyzed HIV-1 viral load GWAS data from three diverse populations. Our analysis pinpointed the MHC association to three amino acid positions (97, 67 and 156) marking three consecutive pockets within the HLA-B peptide binding groove, explaining 12.9% of trait variance, and obviating effects of previously reported associations from population-specific HIV studies.