Benchmarking GCC 4.2 Through GCC 4.8 On AMD & Intel Hardware
The Fortran-based NASA Parallel Benchmark with OpenMP only worked going back to the GCC 4.4 series, but even going back with just the past five releases was quite interesting. For both the Intel Core i7 and AMD Opteron systems, the computationally-intensive NPB LU.A benchmark performed much better with each succeeding GCC release.
The LAMMPS molecular dynamics simulator was largely unchanged by more recent versions of GCC on the Intel and AMD 64-bit systems.
The FFTE performance improves with more recent versions of GCC.