Past urban research on Western nations tends to show high suicide rates in inner city and socioeconomically deprived areas. However, little is known about geographic variations in suicide in non-Western cities. We used Bayesian hierarchical models to estimate smoothed standardised mortality ratios (2005-2010) for suicide in people aged 10 years or above in each geographic unit in Hong Kong at two levels, i.e. large street block (n = 1639; median population = 1860) and small tertiary planning unit group (n = 204; median population = 14,850). We further analysed their associations with a range of area socioeconomic characteristics and a deprivation index. The "city centre" of Hong Kong, a generally non-deprived area, showed mostly below average suicide rates. However, there were high rates concentrating in some socioeconomically deprived, densely populated areas, including some inner city areas, across the city. Males had greater geographic variations in rates than females, except the elderly group. The use of smaller geographic units revealed finer detailed suicide distribution than the use of larger units, and showed that suicide rates were associated with indicators of socioeconomic deprivation (population with non-professional jobs and low median household income), and social fragmentation (proportions of unmarried adults and divorced/separated adults), but not with Gini coefficient. Sex/age groups had different associations with suicide rates. Areas in the most deprived quintile had a suicide rate more than two times higher than the least deprived. The association between suicide and deprivation was stronger in males than females and more marked in the younger populations compared to the elderly. The spatial distribution of suicide in Hong Kong showed distinct patterning and a stronger association with income compared to findings from Western countries. Suicide prevention strategies should consider tackling the marked socioeconomic gradient in suicide and high risk in young and middle-aged males living in deprived areas.