Fig. 7From: Scalable non-negative matrix tri-factorizationEfficiency of multi-GPU architectures for different block configurations. Efficiency is represented by the fraction of linear speedup. The number of GPU devices is equal to the number of matrix blocksBack to article page