Traditional smallholder farming systems are characterized by low yields and high risks of crop failure and food insecurity. Through a biophysical model, PARCHED-THIRST and a socio-economic farming systems simulation model, OLYMPE, we evaluated the performance of farming practices based on maize yield, gross margin and total family balance over a 10-year period in semi-arid Olifants River Basin of South Africa. Farm profitability under scenarios of different maize productions, maize grain and fertiliser price variations were explored for the identified farming systems. Farm types (A to E) were identified from farm surveys, and validated with farmers and extension officers. The order of vulnerability to severe droughts and food insecurity, starting with the most vulnerable is farm Type B, C, D, A and E. Severe drought or flood shock resulted in highest farm gross margin and total family balance reductions, partly due to loss of production for family consumption. Labour returns ranged from US$ 62/capita.year for crop-based farm types to US$ 363/capita.year for livestock-based farm Type E. Results revealed that livestock and crop diversification are most proficient strategies to ensure stable income and food security for smallholder farmers. Thus, smallholder farming technology innovations and policies should engage in solutions to poor yields and livestock farming.