Algebraic multigrid (AMG) is a popular solver for large-scale scientific computing and an essential component of many simulation codes and has shown to be extremely efficient on distributed-memory architectures. Past performance studies have shown that its performance and scalability depend significantly on the specific architecture of a computer including node architecture, interconnect, and operating system capabilities. We present scaling studies of several AMG variants on Blue Gene/Q. The effect of various MPI/OpenMP combinations, different smoothers and the use of multiplicative and additive V-cycles are investigated.