SPEC CPU result comparison

15th October 2026 - Normalized report

SPEC2017 FPrate: PGO+LTO generic -O2 - Released GCCs vs Trunk

-O2 -flto -fprofile-use

AMD EPYC 8635P (zen5 based Sorano)

Run-times

Benchmark GCC 15 GCC 16 Trunk GCC 16 Trunk
503.bwaves_r 100.00% 100.00% 99.25% 100.00% 99.25%
507.cactuBSSN_r 100.00% 97.68% 99.13% 100.00% 101.48%
508.namd_r 100.00% 100.00% 100.82% 100.00% 100.82%
510.parest_r 100.00% 100.00% 92.35% 100.00% 92.35%
511.povray_r 100.00% 98.64% 98.64% 100.00% 100.00%
519.lbm_r 100.00% 100.00% 100.00% 100.00% 100.00%
521.wrf_r 100.00% 98.77% 100.93% 100.00% 102.19%
526.blender_r 100.00% 98.68% 108.77% 100.00% 110.22%
527.cam4_r 100.00% 101.71% 99.02% 100.00% 97.36%
538.imagick_r 100.00% 76.84% 101.64% 100.00% 132.27%
544.nab_r 100.00% 100.00% 89.69% 100.00% 89.69%
549.fotonik3d_r 100.00% 100.00% 101.68% 100.00% 101.68%
554.roms_r 100.00% 98.47% 99.08% 100.00% 100.62%
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% 103.95% 103.59% 100.00% 99.65%
507.cactuBSSN_r 100.00% 103.52% 98.36% 100.00% 95.01%
508.namd_r 100.00% 100.41% 100.66% 100.00% 100.24%
510.parest_r 100.00% 101.33% 101.53% 100.00% 100.20%
511.povray_r 100.00% 103.24% 101.53% 100.00% 98.34%
519.lbm_r 100.00% 100.10% 100.72% 100.00% 100.61%
521.wrf_r 100.00% 102.04% 101.87% 100.00% 99.83%
526.blender_r 100.00% 100.46% 100.45% 100.00% 100.00%
527.cam4_r 100.00% 100.25% 100.19% 100.00% 99.94%
538.imagick_r 100.00% 102.76% 102.21% 100.00% 99.47%
544.nab_r 100.00% 100.58% 100.50% 100.00% 99.92%
549.fotonik3d_r 100.00% 105.38% 104.50% 100.00% 99.16%
554.roms_r 100.00% 101.72% 99.21% 100.00% 97.54%
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% 99.47% 99.47% 100.00% 100.00%
507.cactuBSSN_r 100.00% 97.97% 99.10% 100.00% 101.15%
508.namd_r 100.00% 100.00% 99.39% 100.00% 99.39%
510.parest_r 100.00% 100.74% 99.75% 100.00% 99.02%
511.povray_r 100.00% 96.72% 97.89% 100.00% 101.21%
519.lbm_r 100.00% 100.00% 100.00% 100.00% 100.00%
521.wrf_r 100.00% 99.14% 100.65% 100.00% 101.52%
526.blender_r 100.00% 99.39% 98.98% 100.00% 99.59%
527.cam4_r 100.00% 101.76% 99.34% 100.00% 97.62%
538.imagick_r 100.00% 86.63% 101.65% 100.00% 117.34%
544.nab_r 100.00% 100.00% 93.62% 100.00% 93.62%
549.fotonik3d_r 100.00% 100.89% 100.00% 100.00% 99.12%
554.roms_r 100.00% 100.00% 100.00% 100.00% 100.00%
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% 103.95% 103.59% 100.00% 99.65%
507.cactuBSSN_r 100.00% 103.51% 98.35% 100.00% 95.01%
508.namd_r 100.00% 100.40% 100.65% 100.00% 100.24%
510.parest_r 100.00% 101.33% 101.52% 100.00% 100.20%
511.povray_r 100.00% 103.25% 101.54% 100.00% 98.34%
519.lbm_r 100.00% 100.10% 100.71% 100.00% 100.61%
521.wrf_r 100.00% 102.04% 101.87% 100.00% 99.83%
526.blender_r 100.00% 100.46% 100.45% 100.00% 100.00%
527.cam4_r 100.00% 100.25% 100.19% 100.00% 99.94%
538.imagick_r 100.00% 102.75% 102.21% 100.00% 99.47%
544.nab_r 100.00% 100.58% 100.50% 100.00% 99.92%
549.fotonik3d_r 100.00% 105.40% 104.51% 100.00% 99.16%
554.roms_r 100.00% 101.72% 99.21% 100.00% 97.53%
Smaller binary sizes are always better.