Gurobi 3.0 Wins Performance Benchmarks

In April 2010, Professor Hans Mittelmann updated his benchmark results for optimization software. For linear programming, he reported results for the dual simplex algorithm showing that Gurobi Optimizer 3.0 is 1.4 times faster than competing systems. Also, he reported results for the barrier algorithm on four processing cores showing Gurobi Optimizer 3.0 is 1.9 times faster than competing systems. For integer programming, he reported results for a single processing core showing that Gurobi Optimizer 3.0 is 1.8 times faster than competing systems in time to proven optimality. Using parallel processing for integer programming, he reported results showing that Gurobi Optimizer 3.0 is 2.4 times faster than competing systems on six cores and 2.8 times faster than competing systems on twelve cores.  This shows that Gurobi Optimizer takes better advantage of parallel processing than competing systems. The time to find the first feasible solution showed an even greater advantage for Gurobi: Gurobi was 4.8 times faster than competing systems. And for many users, the time to find the first feasible solution is often more useful than the time to find the optimal solution. Here are links to the detailed results:

• Benchmark of parallel LP solvers
• Mixed Integer Linear Programming Benchmark (serial codes)
• Mixed Integer Linear Programming Benchmark (parallel codes)
• Feasibility Benchmark