IBM ILOG CPLEX Optimizer confirms its performance leadership with the Version 12.8 CPLEX Optimizer which provides breakthrough improvements in performance.
Mixed integer programming models with linear terms are solved by the CPLEX MIP Optimizer 23% faster on average for large problems (>1000s) than version 12.7. For models with convex quadratic terms, improvements on larger models (>10s) is 13% on average. On a special model type called QDP/SOCP the speedup is between 5-10% on average.
CPLEX MILP performance evolution
CPLEX 12.7.0 vs 12.8.0: LP performance improvement
CPLEX 12.7.0 vs 12.8.0: MILP performance improvement
CPLEX 12.7.0 vs 12.8.0: MIQP performance improvement
CPLEX 12.7.0 vs 12.8.0:QCP/SOCP performance improvement
CPLEX CP Optimizer
Test-cases are separated into two groups:
Integer problems, which include rostering, matching, sports scheduling problems.
Scheduling problems, which include resource-constrained and job-shop scheduling problems.
On performance of purely combinatorial/integer problems a breakthrough speed improvement has been achieved with a staggering +2.5x, i.e. a more than 150% faster solve. On purely scheduling problems with COS 12.8, we have achieved a 10% speedup compared to 12.7.1. Compared to 12.7.0, a 50% speed up can be observed.