SPEC CPU result comparison

15th October 2026 - Normalized report

SPEC2017 FPrate: LTO -Ofast native - LLVM and GCC

-Ofast -march=native -flto

AMD EPYC 8635P (zen5 based Sorano)

Run-times

Benchmark LLVM GCC
503.bwaves_r 100.00% 81.00%
507.cactuBSSN_r 100.00% 96.99%
508.namd_r 100.00% 107.57%
510.parest_r 100.00% 80.77%
511.povray_r 100.00% 96.62%
519.lbm_r 100.00% 49.65%
521.wrf_r 100.00% 97.06%
526.blender_r 100.00% 100.72%
527.cam4_r 100.00% 78.04%
538.imagick_r 100.00% 52.75%
544.nab_r 100.00% 104.35%
549.fotonik3d_r 100.00% 100.67%
554.roms_r 100.00% 93.43%
LLVM clang version 23.1.3 (/labs/mjambor/git/llvm 0d261d1ca552c95a8f007e061c787ac7132fbcbc), flang version 23.1.3 (/labs/mjambor/git/llvm 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% 4.44%
507.cactuBSSN_r 100.00% 97.59%
508.namd_r 100.00% 122.83%
510.parest_r 100.00% 90.52%
511.povray_r 100.00% 90.40%
519.lbm_r 100.00% 52.10%
521.wrf_r 100.00% 96.19%
526.blender_r 100.00% 79.61%
527.cam4_r 100.00% 100.81%
538.imagick_r 100.00% 121.31%
544.nab_r 100.00% 102.04%
549.fotonik3d_r 100.00% 34.80%
554.roms_r 100.00% 58.44%
Smaller binary sizes are always better.

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

Run-times

Benchmark LLVM GCC
503.bwaves_r 100.00% 84.13%
507.cactuBSSN_r 100.00% 91.98%
508.namd_r 100.00% 105.67%
510.parest_r 100.00% 95.79%
511.povray_r 100.00% 97.79%
519.lbm_r 100.00% 160.34%
521.wrf_r 100.00% 95.24%
526.blender_r 100.00% 111.49%
527.cam4_r 100.00% 82.57%
538.imagick_r 100.00% 69.01%
544.nab_r 100.00% 106.55%
549.fotonik3d_r 100.00% 100.47%
554.roms_r 100.00% 92.67%
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% 4.64%
507.cactuBSSN_r 100.00% 95.31%
508.namd_r 100.00% 135.48%
510.parest_r 100.00% 93.41%
511.povray_r 100.00% 96.09%
519.lbm_r 100.00% 71.09%
521.wrf_r 100.00% 101.01%
526.blender_r 100.00% 81.83%
527.cam4_r 100.00% 101.49%
538.imagick_r 100.00% 115.32%
544.nab_r 100.00% 101.50%
549.fotonik3d_r 100.00% 34.94%
554.roms_r 100.00% 59.13%
Smaller binary sizes are always better.