SPEC CPU result comparison

15th October 2026 - Normalized report

SPEC2017 FPrate: Trunk LTO -O2 - generic vs x86-64-v3 vs native

-O2 -flto vs -O2 -flto and a given -march

AMD EPYC 8635P (zen5 based Sorano)

Run-times

Benchmark Generic x86-64-v3 Native x86-64-v3 Native
503.bwaves_r 100.00% 99.25% 99.25% 100.00% 100.00%
507.cactuBSSN_r 100.00% 86.67% 82.61% 100.00% 95.32%
508.namd_r 100.00% 89.69% 84.74% 100.00% 94.48%
510.parest_r 100.00% 100.00% 99.73% 100.00% 99.73%
511.povray_r 100.00% 98.38% 99.54% 100.00% 101.17%
519.lbm_r 100.00% 84.70% 79.00% 100.00% 93.28%
521.wrf_r 100.00% 91.84% 90.38% 100.00% 98.41%
526.blender_r 100.00% 100.00% 99.32% 100.00% 99.32%
527.cam4_r 100.00% 87.78% 90.95% 100.00% 103.61%
538.imagick_r 100.00% 61.02% 62.08% 100.00% 101.74%
544.nab_r 100.00% 92.97% 92.97% 100.00% 100.00%
549.fotonik3d_r 100.00% 96.18% 96.18% 100.00% 100.00%
554.roms_r 100.00% 91.82% 91.29% 100.00% 99.43%
The compiler used was GCC 17.0.0 20261015 (experimental) (revision 84348add236100e5f69a3e9ec4f7b1038bf2b83d).
Naturally, the smaller the run-times, the better.
See also continuous gcc tuning benchmarking at https://lnt.opensuse.org/db_default/v4/SPEC/spec_report/tuning.

Non-debug executable sizes

Benchmark Generic x86-64-v3 Native x86-64-v3 Native
503.bwaves_r 100.00% 89.57% 94.20% 100.00% 105.17%
507.cactuBSSN_r 100.00% 97.56% 116.86% 100.00% 119.78%
508.namd_r 100.00% 94.67% 99.78% 100.00% 105.40%
510.parest_r 100.00% 99.75% 100.88% 100.00% 101.14%
511.povray_r 100.00% 97.47% 98.11% 100.00% 100.66%
519.lbm_r 100.00% 94.14% 92.93% 100.00% 98.72%
521.wrf_r 100.00% 99.06% 99.15% 100.00% 100.09%
526.blender_r 100.00% 99.33% 99.82% 100.00% 100.49%
527.cam4_r 100.00% 99.77% 99.89% 100.00% 100.12%
538.imagick_r 100.00% 98.94% 97.94% 100.00% 98.98%
544.nab_r 100.00% 99.26% 99.60% 100.00% 100.34%
549.fotonik3d_r 100.00% 100.72% 102.33% 100.00% 101.60%
554.roms_r 100.00% 100.26% 99.36% 100.00% 99.11%
Smaller binary sizes are always better.

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

Run-times

Benchmark Generic x86-64-v3
503.bwaves_r 100.00% 83.54%
507.cactuBSSN_r 100.00% 94.08%
508.namd_r 100.00% 86.42%
510.parest_r 100.00% 101.41%
511.povray_r 100.00% 96.65%
519.lbm_r 100.00% 92.57%
521.wrf_r 100.00% 90.32%
526.blender_r 100.00% 98.96%
527.cam4_r 100.00% 86.78%
538.imagick_r 100.00% 65.15%
544.nab_r 100.00% 93.55%
549.fotonik3d_r 100.00% 96.07%
554.roms_r 100.00% 91.48%
The compiler used was GCC 17.0.0 20261015 (experimental) (revision 84348add236100e5f69a3e9ec4f7b1038bf2b83d).
Naturally, the smaller the run-times, the better.
See also continuous gcc tuning benchmarking at https://lnt.opensuse.org/db_default/v4/SPEC/spec_report/tuning.

Non-debug executable sizes

Benchmark Generic x86-64-v3
503.bwaves_r 100.00% 89.59%
507.cactuBSSN_r 100.00% 97.56%
508.namd_r 100.00% 94.67%
510.parest_r 100.00% 99.75%
511.povray_r 100.00% 97.47%
519.lbm_r 100.00% 94.39%
521.wrf_r 100.00% 99.06%
526.blender_r 100.00% 99.33%
527.cam4_r 100.00% 99.77%
538.imagick_r 100.00% 98.94%
544.nab_r 100.00% 99.26%
549.fotonik3d_r 100.00% 100.72%
554.roms_r 100.00% 100.26%
Smaller binary sizes are always better.