SPEC CPU result comparison

15th October 2026 - Normalized report

SPEC2017 FPrate: Standard -Ofast native - LLVM and GCC

-Ofast -march=native -mtune=native

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

Run-times

Benchmark LLVM GCC
503.bwaves_r 100.00% 56.54%
507.cactuBSSN_r 100.00% 90.98%
508.namd_r 100.00% 104.89%
510.parest_r 100.00% 84.87%
511.povray_r 100.00% 97.17%
519.lbm_r 100.00% 150.21%
521.wrf_r 100.00% 95.26%
526.blender_r 100.00% 107.35%
527.cam4_r 100.00% 82.33%
538.imagick_r 100.00% 63.35%
544.nab_r 100.00% 103.27%
549.fotonik3d_r 100.00% 101.42%
554.roms_r 100.00% 91.50%
LLVM clang version 23.1.3 (https://github.com/llvm/llvm-project.git 0d261d1ca552c95a8f007e061c787ac7132fbcbc), flang version 23.1.3 (https://github.com/llvm/llvm-project.git 0d261d1ca552c95a8f007e061c787ac7132fbcbc)
GCC GCC 17.0.0 20261015 (experimental) (revision 84348add236100e5f69a3e9ec4f7b1038bf2b83d)
In the case of LLVM -fno-finite-math-only was used to compile 511.povray_r (that benchmark only) to avoid a known issue.
Naturally, the smaller the run-times, the better.

Non-debug executable sizes

Benchmark LLVM GCC
503.bwaves_r 100.00% 4.55%
507.cactuBSSN_r 100.00% 102.08%
508.namd_r 100.00% 116.66%
510.parest_r 100.00% 99.83%
511.povray_r 100.00% 109.25%
519.lbm_r 100.00% 68.82%
521.wrf_r 100.00% 100.32%
526.blender_r 100.00% 86.56%
527.cam4_r 100.00% 239.24%
538.imagick_r 100.00% 142.37%
544.nab_r 100.00% 99.33%
549.fotonik3d_r 100.00% 37.22%
554.roms_r 100.00% 63.69%
Smaller binary sizes are always better.