Enterprise Optimization Algorithms Scale to Massive ComplexityNVIDIA cuOpt mPDLP scales linear programming to 100M+ variablesTechnology Services › Scientific Research and Development Services › Research and Development in the Physical, Engineering, and Life SciencesDataNVIDIA introduced mPDLP, a multi-GPU solver for linear programming problems that distributes optimization across NVLink-connected GPUs, achieving nearly 10x faster solve times on benchmark problems and up to 6x lower peak memory usage per GPU. For supply chain and energy-grid planning teams, this breakthrough removes the single-GPU bottleneck that historically forced large optimization problems to take hours to converge, enabling faster decision cycles in real-time planning workflows.Oct 8, 2026Read full article →
