SPEC CPU result comparison

15th October 2026 - Normalized report

SPEC2017 FPrate: Trunk LTO -Ofast native - using 128/256/512bit vectors vs default

-Ofast -march=native -mtune=native with and without -mprefer-vector-width=128,256 and/or 512

AMD EPYC 8635P (zen5 based Sorano)

Run-times

Benchmark 128 256 Default 256 Default
503.bwaves_r 100.00% 99.44% 100.00% 100.00% 100.56%
507.cactuBSSN_r 100.00% 100.00% 100.00% 100.00% 100.00%
508.namd_r 100.00% 98.99% 100.00% 100.00% 101.02%
510.parest_r 100.00% 95.49% 94.67% 100.00% 99.14%
511.povray_r 100.00% 102.26% 100.50% 100.00% 98.28%
519.lbm_r 100.00% 101.42% 101.42% 100.00% 100.00%
521.wrf_r 100.00% 85.82% 84.00% 100.00% 97.88%
526.blender_r 100.00% 100.97% 101.69% 100.00% 100.72%
527.cam4_r 100.00% 91.75% 92.53% 100.00% 100.84%
538.imagick_r 100.00% 97.88% 86.24% 100.00% 88.11%
544.nab_r 100.00% 100.84% 100.84% 100.00% 100.00%
549.fotonik3d_r 100.00% 103.91% 98.48% 100.00% 94.77%
554.roms_r 100.00% 79.42% 71.84% 100.00% 90.45%
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 128 256 Default 256 Default
503.bwaves_r 100.00% 104.17% 108.32% 100.00% 103.99%
507.cactuBSSN_r 100.00% 102.88% 112.75% 100.00% 109.60%
508.namd_r 100.00% 106.04% 111.23% 100.00% 104.89%
510.parest_r 100.00% 103.81% 106.81% 100.00% 102.89%
511.povray_r 100.00% 105.38% 106.76% 100.00% 101.31%
519.lbm_r 100.00% 101.46% 102.06% 100.00% 100.60%
521.wrf_r 100.00% 101.48% 103.07% 100.00% 101.57%
526.blender_r 100.00% 96.82% 98.69% 100.00% 101.94%
527.cam4_r 100.00% 100.40% 100.83% 100.00% 100.42%
538.imagick_r 100.00% 104.89% 112.56% 100.00% 107.32%
544.nab_r 100.00% 101.05% 102.35% 100.00% 101.29%
549.fotonik3d_r 100.00% 116.90% 124.57% 100.00% 106.55%
554.roms_r 100.00% 105.49% 108.86% 100.00% 103.20%
Smaller binary sizes are always better.

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

Run-times

Benchmark 128 Default
503.bwaves_r 100.00% 98.51%
507.cactuBSSN_r 100.00% 100.55%
508.namd_r 100.00% 100.00%
510.parest_r 100.00% 96.90%
511.povray_r 100.00% 99.33%
519.lbm_r 100.00% 99.20%
521.wrf_r 100.00% 86.51%
526.blender_r 100.00% 100.00%
527.cam4_r 100.00% 91.33%
538.imagick_r 100.00% 102.37%
544.nab_r 100.00% 98.68%
549.fotonik3d_r 100.00% 98.62%
554.roms_r 100.00% 83.99%
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 128 Default
503.bwaves_r 100.00% 107.71%
507.cactuBSSN_r 100.00% 105.01%
508.namd_r 100.00% 108.90%
510.parest_r 100.00% 104.39%
511.povray_r 100.00% 104.16%
519.lbm_r 100.00% 101.14%
521.wrf_r 100.00% 102.75%
526.blender_r 100.00% 97.09%
527.cam4_r 100.00% 100.65%
538.imagick_r 100.00% 107.04%
544.nab_r 100.00% 101.28%
549.fotonik3d_r 100.00% 119.96%
554.roms_r 100.00% 109.54%
Smaller binary sizes are always better.