SPEC CPU result comparison

15th October 2026 - Normalized report

SPECFP 2006: Trunk generic -O2 - standard vs LTO vs LTO+PGO

-O2

AMD EPYC 8635P (zen5 based Sorano)

Run-times

Benchmark LTO LTO+PGO
410.bwaves 100.00% 98.87%
416.gamess 100.00% 104.00%
433.milc 100.00% 50.30%
434.zeusmp 100.00% 98.90%
435.gromacs 100.00% 98.38%
436.cactusADM 100.00% 101.06%
437.leslie3d 100.00% 97.46%
444.namd 100.00% 99.39%
447.dealII 100.00% 101.28%
450.soplex 100.00% 100.00%
453.povray 100.00% 79.44%
454.calculix 100.00% 106.64%
459.GemsFDTD 100.00% 99.10%
465.tonto 100.00% 92.23%
470.lbm 100.00% 78.44%
481.wrf 100.00% 94.17%
482.sphinx3 100.00% 100.00%
The compiler used was GCC 17.0.0 20261015 (experimental) (revision 84348add236100e5f69a3e9ec4f7b1038bf2b83d).
Naturally, the smaller the run-times, the better.
See also continuous gcc options benchmarking at https://lnt.opensuse.org/db_default/v4/SPEC/spec_report/options.

Non-debug executable sizes

Benchmark LTO LTO+PGO
410.bwaves 100.00% 117.45%
416.gamess 100.00% 99.79%
433.milc 100.00% 106.29%
434.zeusmp 100.00% 100.00%
435.gromacs 100.00% 89.29%
436.cactusADM 100.00% 81.40%
437.leslie3d 100.00% 95.37%
444.namd 100.00% 77.59%
447.dealII 100.00% 79.49%
450.soplex 100.00% 82.40%
453.povray 100.00% 83.37%
454.calculix 100.00% 86.92%
459.GemsFDTD 100.00% 85.61%
465.tonto 100.00% 90.33%
470.lbm 100.00% 101.75%
481.wrf 100.00% 86.24%
482.sphinx3 100.00% 91.29%
Smaller binary sizes are always better.

AMD EPYC 75F3 (zen3 based high-frequency Milan)

Run-times

Benchmark LTO LTO+PGO
410.bwaves 100.00% 91.50%
416.gamess 100.00% 109.96%
433.milc 100.00% 38.35%
434.zeusmp 100.00% 95.84%
435.gromacs 100.00% 97.17%
436.cactusADM 100.00% 103.02%
437.leslie3d 100.00% 95.33%
444.namd 100.00% 100.00%
447.dealII 100.00% 103.35%
450.soplex 100.00% 99.13%
453.povray 100.00% 86.83%
454.calculix 100.00% 104.85%
459.GemsFDTD 100.00% 100.00%
465.tonto 100.00% 94.64%
470.lbm 100.00% 93.61%
481.wrf 100.00% 100.00%
482.sphinx3 100.00% 99.15%
The compiler used was GCC 17.0.0 20261015 (experimental) (revision 84348add236100e5f69a3e9ec4f7b1038bf2b83d).
Naturally, the smaller the run-times, the better.
See also continuous gcc options benchmarking at https://lnt.opensuse.org/db_default/v4/SPEC/spec_report/options.

Non-debug executable sizes

Benchmark LTO LTO+PGO
410.bwaves 100.00% 117.25%
416.gamess 100.00% 99.79%
433.milc 100.00% 106.29%
434.zeusmp 100.00% 100.00%
435.gromacs 100.00% 89.30%
436.cactusADM 100.00% 81.43%
437.leslie3d 100.00% 95.40%
444.namd 100.00% 77.59%
447.dealII 100.00% 79.48%
450.soplex 100.00% 82.40%
453.povray 100.00% 83.37%
454.calculix 100.00% 86.92%
459.GemsFDTD 100.00% 85.60%
465.tonto 100.00% 90.34%
470.lbm 100.00% 101.74%
481.wrf 100.00% 86.24%
482.sphinx3 100.00% 91.30%
Smaller binary sizes are always better.