This is from an excellent paper by Rumman Khan, Oliver Morrissey and Paul Mosley:

Between 1990 and 2012, for most of the developing world, poverty has halved or more than halved except in sub-Saharan Africa (SSA). The simple poverty headcount fell from about 60% to 15% in East Asia; 50% to 25% in South Asia; 20% to 10% in Latin America; but only from 57% to under 43% in SSA (Beegle, Christiaensen, Dabalen and Gaddis, 2016: 21- 22). This is despite more than a decade of impressive growth in SSA, averaging 5-6 per cent per year since the late 1990s (Devarajan, 2013: S9). Some countries did (almost) halve poverty, such as Ghana (McKay and Osei-Assibey, 2017) and Uganda (Kakande, 2010), and many achieved significant reductions. In contrast, populous countries such as South Africa and Nigeria, on the available evidence, have not achieved significant poverty reduction.

The authors note that the effects of growth on poverty reduction across Africa has been bimodal. And this is their explanation:

To explain variation within SSA in poverty reduction, we consider aspects of colonial experience associated with the emergence of differing potential for redistributive policies to emerge after independence. Following the approach of Myint (1976) and others, we classify SSA countries into two groups according to the economic strategies used by the colonial authorities, using pre-independence data on factors such as inequality, land ownership by Europeans and political participation by Africans (the process is detailed in Appendix A, with validation by cluster analysis). In smallholder production economies, African agricultural smallholders had economic and some political participation. In contrast, extractive production economies dominated by foreign-owned mines and large-scale farms fostered the emergence of an elite politics characterised by urban bias and capital-intensive production technologies. During the colonial period African economies became clustered around a bimodal structure, which provided better opportunities to the poor in countries whose production was based on the development of labour-intensive smallholder exports than in countries whose growth strategy was based more on capital-intensive mines and large farms. We then test if the growth elasticity of poverty differs between these two groups of countries, using available (PovcalNet) poverty data since 1985, noting that mean growth rates for the two groups were very similar. The analysis shows that the smallholder group significantly outperformed the extractive group, smallholder experience is a significant predictor of poverty reduction, and inclusion of other potential explanatory variables does not alter the conclusion.

I recommend you read the whole paper (including the very rich appendix).