SPEC CPU result comparison

15th October 2026 - Normalized report

SPEC2017 FPrate: LTO x86-64-v3 -O3 - Released GCCs vs Trunk

-O3 -flto -march=x86-64-v3

AMD EPYC 8635P (zen5 based Sorano)

Run-times

Benchmark GCC 15 GCC 16 Trunk GCC 16 Trunk
503.bwaves_r 100.00% 101.26% 100.00% 100.00% 98.76%
507.cactuBSSN_r 100.00% 102.91% 100.00% 100.00% 97.17%
508.namd_r 100.00% 100.00% 99.08% 100.00% 99.08%
510.parest_r 100.00% 98.83% 91.52% 100.00% 92.60%
511.povray_r 100.00% 99.49% 100.51% 100.00% 101.02%
519.lbm_r 100.00% 101.26% 100.00% 100.00% 98.76%
521.wrf_r 100.00% 97.60% 98.97% 100.00% 101.40%
526.blender_r 100.00% 100.00% 99.30% 100.00% 99.30%
527.cam4_r 100.00% 100.00% 100.00% 100.00% 100.00%
538.imagick_r 100.00% 100.00% 101.85% 100.00% 101.85%
544.nab_r 100.00% 100.00% 89.06% 100.00% 89.06%
549.fotonik3d_r 100.00% 99.59% 99.59% 100.00% 100.00%
554.roms_r 100.00% 100.00% 98.39% 100.00% 98.39%
GCC 15 GCC 15.3.0 (revision 4db0e8df15bef836558857c291c323add11d035c)
GCC 16 GCC 16.2.0 (revision 78d4ac73dd391005b895a6148cd9831e28e1208b)
Trunk GCC 17.0.0 20261015 (experimental) (revision 84348add236100e5f69a3e9ec4f7b1038bf2b83d)
Naturally, the smaller the run-times, the better.
See also continuous gcc branch benchmarking at https://lnt.opensuse.org/db_default/v4/SPEC/spec_report/branch.

Non-debug executable sizes

Benchmark GCC 15 GCC 16 Trunk GCC 16 Trunk
503.bwaves_r 100.00% 101.28% 101.32% 100.00% 100.03%
507.cactuBSSN_r 100.00% 103.24% 100.28% 100.00% 97.14%
508.namd_r 100.00% 100.26% 96.12% 100.00% 95.87%
510.parest_r 100.00% 108.56% 106.59% 100.00% 98.19%
511.povray_r 100.00% 102.57% 100.18% 100.00% 97.67%
519.lbm_r 100.00% 100.00% 98.10% 100.00% 98.10%
521.wrf_r 100.00% 100.53% 99.56% 100.00% 99.03%
526.blender_r 100.00% 100.14% 98.60% 100.00% 98.46%
527.cam4_r 100.00% 100.28% 100.21% 100.00% 99.93%
538.imagick_r 100.00% 102.36% 97.40% 100.00% 95.16%
544.nab_r 100.00% 100.87% 100.46% 100.00% 99.59%
549.fotonik3d_r 100.00% 102.16% 100.92% 100.00% 98.79%
554.roms_r 100.00% 104.86% 97.88% 100.00% 93.34%
Smaller binary sizes are always better.

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

Run-times

Benchmark GCC 15 GCC 16 Trunk GCC 16 Trunk
503.bwaves_r 100.00% 100.60% 100.00% 100.00% 99.41%
507.cactuBSSN_r 100.00% 100.00% 100.00% 100.00% 100.00%
508.namd_r 100.00% 100.00% 100.72% 100.00% 100.72%
510.parest_r 100.00% 99.50% 96.26% 100.00% 96.74%
511.povray_r 100.00% 102.55% 99.54% 100.00% 97.06%
519.lbm_r 100.00% 100.68% 100.00% 100.00% 99.33%
521.wrf_r 100.00% 99.02% 100.73% 100.00% 101.73%
526.blender_r 100.00% 98.94% 97.25% 100.00% 98.29%
527.cam4_r 100.00% 100.74% 100.74% 100.00% 100.00%
538.imagick_r 100.00% 100.34% 101.38% 100.00% 101.03%
544.nab_r 100.00% 100.00% 92.47% 100.00% 92.47%
549.fotonik3d_r 100.00% 99.54% 100.46% 100.00% 100.93%
554.roms_r 100.00% 98.69% 98.04% 100.00% 99.34%
GCC 15 GCC 15.3.0 (revision 4db0e8df15bef836558857c291c323add11d035c)
GCC 16 GCC 16.2.0 (revision 78d4ac73dd391005b895a6148cd9831e28e1208b)
Trunk GCC 17.0.0 20261015 (experimental) (revision 84348add236100e5f69a3e9ec4f7b1038bf2b83d)
Naturally, the smaller the run-times, the better.
See also continuous gcc branch benchmarking at https://lnt.opensuse.org/db_default/v4/SPEC/spec_report/branch.

Non-debug executable sizes

Benchmark GCC 15 GCC 16 Trunk GCC 16 Trunk
503.bwaves_r 100.00% 101.28% 101.37% 100.00% 100.09%
507.cactuBSSN_r 100.00% 103.24% 100.28% 100.00% 97.14%
508.namd_r 100.00% 100.26% 96.12% 100.00% 95.87%
510.parest_r 100.00% 108.56% 106.59% 100.00% 98.19%
511.povray_r 100.00% 102.56% 100.18% 100.00% 97.67%
519.lbm_r 100.00% 100.47% 98.58% 100.00% 98.11%
521.wrf_r 100.00% 100.53% 99.56% 100.00% 99.03%
526.blender_r 100.00% 100.14% 98.61% 100.00% 98.47%
527.cam4_r 100.00% 100.28% 100.21% 100.00% 99.93%
538.imagick_r 100.00% 102.36% 97.40% 100.00% 95.16%
544.nab_r 100.00% 100.85% 100.46% 100.00% 99.61%
549.fotonik3d_r 100.00% 102.16% 100.92% 100.00% 98.79%
554.roms_r 100.00% 104.86% 97.88% 100.00% 93.34%
Smaller binary sizes are always better.