SPEC CPU result comparison

15th October 2026 - Normalized report

SPEC2026 FPrate: Standard -Ofast native - LLVM and GCC

-Ofast -march=native -mtune=native

Please note that event though the header of this comparison says -Ofast, benchmark 767.nest_r was additionally compiled with -fno-finite-math-only. Furthermore, benchmarks 737.gmsh_r and 748.flightdm_r were compiled with -fno-fast-math, so a big chunk of -Ofast was in fact disabled. (This might put GCC at a disadvantage compared to LLVM-based compilers because the latter defaults to -fno-math-errno and -fno-trapping-math, in future measurements, we will try to enable these flags.)

AMD EPYC 8635P (zen5 based Sorano)

Run-times

Benchmark LLVM GCC
709.cactus_r 100.00% 106.43%
722.palm_r 100.00% 114.05%
731.astcenc_r 100.00% 116.03%
736.ocio_r 100.00% 93.03%
737.gmsh_r 100.00% 93.58%
748.flightdm_r 100.00% 91.04%
749.fotonik3d_r 100.00% 101.34%
765.roms_r 100.00% 92.92%
766.femflow_r 100.00% 82.14%
767.nest_r 100.00% 90.00%
772.marian_r 100.00% 60.00%
782.lbm_r 100.00% 61.18%
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)
Naturally, the smaller the run-times, the better.

Non-debug executable sizes

Benchmark LLVM GCC
709.cactus_r 100.00% 106.65%
722.palm_r 100.00% 77.29%
731.astcenc_r 100.00% 137.87%
736.ocio_r 100.00% 121.76%
737.gmsh_r 100.00% 107.35%
748.flightdm_r 100.00% 119.86%
749.fotonik3d_r 100.00% 38.12%
765.roms_r 100.00% 68.46%
766.femflow_r 100.00% 87.99%
767.nest_r 100.00% 110.21%
772.marian_r 100.00% 108.35%
782.lbm_r 100.00% 52.40%
Smaller binary sizes are always better.

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

Run-times

Benchmark LLVM GCC
709.cactus_r 100.00% 99.15%
722.palm_r 100.00% 114.24%
731.astcenc_r 100.00% 102.40%
736.ocio_r 100.00% 94.78%
737.gmsh_r 100.00% 95.44%
748.flightdm_r 100.00% 94.24%
749.fotonik3d_r 100.00% 102.65%
765.roms_r 100.00% 90.82%
766.femflow_r 100.00% 100.00%
767.nest_r 100.00% 99.23%
772.marian_r 100.00% 107.41%
782.lbm_r 100.00% 110.93%
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)
Naturally, the smaller the run-times, the better.

Non-debug executable sizes

Benchmark LLVM GCC
709.cactus_r 100.00% 108.60%
722.palm_r 100.00% 80.62%
731.astcenc_r 100.00% 157.71%
736.ocio_r 100.00% 127.04%
737.gmsh_r 100.00% 109.03%
748.flightdm_r 100.00% 119.77%
749.fotonik3d_r 100.00% 38.31%
765.roms_r 100.00% 70.83%
766.femflow_r 100.00% 95.04%
767.nest_r 100.00% 109.39%
772.marian_r 100.00% 107.02%
782.lbm_r 100.00% 68.65%
Smaller binary sizes are always better.