SPEC CPU result comparison

15th October 2026 - Normalized report

SPEC2026 FPrate: Trunk LTO -O3 - generic vs x86-64-v3 vs native

-O3 -flto vs -O3 -flto and a given -march

AMD EPYC 8635P (zen5 based Sorano)

Run-times

Benchmark Generic x86-64-v3 Native x86-64-v3 Native
709.cactus_r 100.00% 88.19% 88.19% 100.00% 100.00%
722.palm_r 100.00% 96.89% 84.22% 100.00% 86.93%
731.astcenc_r 100.00% 88.18% 84.03% 100.00% 95.29%
736.ocio_r 100.00% 77.71% 77.07% 100.00% 99.18%
737.gmsh_r 100.00% 100.91% 100.00% 100.00% 99.10%
748.flightdm_r 100.00% 99.01% 101.99% 100.00% 103.01%
749.fotonik3d_r 100.00% 96.48% 88.73% 100.00% 91.97%
765.roms_r 100.00% 80.00% 71.47% 100.00% 89.34%
766.femflow_r 100.00% 56.60% 44.34% 100.00% 78.33%
767.nest_r 100.00% 99.13% 93.72% 100.00% 94.54%
772.marian_r 100.00% 65.52% 41.38% 100.00% 63.16%
782.lbm_r 100.00% 101.76% 95.59% 100.00% 93.94%
The compiler used was GCC 17.0.0 20261015 (experimental) (revision 84348add236100e5f69a3e9ec4f7b1038bf2b83d).
Naturally, the smaller the run-times, the better.
See also continuous gcc tuning benchmarking at https://lnt.opensuse.org/db_default/v4/SPEC/spec_report/tuning.

Non-debug executable sizes

Benchmark Generic x86-64-v3 Native x86-64-v3 Native
709.cactus_r 100.00% 97.70% 98.59% 100.00% 100.91%
722.palm_r 100.00% 103.18% 106.94% 100.00% 103.65%
731.astcenc_r 100.00% 100.81% 105.40% 100.00% 104.55%
736.ocio_r 100.00% 102.45% 104.88% 100.00% 102.37%
737.gmsh_r 100.00% 101.94% 104.21% 100.00% 102.22%
748.flightdm_r 100.00% 99.82% 101.18% 100.00% 101.36%
749.fotonik3d_r 100.00% 116.75% 118.06% 100.00% 101.12%
765.roms_r 100.00% 104.39% 104.08% 100.00% 99.70%
766.femflow_r 100.00% 94.35% 91.95% 100.00% 97.46%
767.nest_r 100.00% 101.18% 102.34% 100.00% 101.15%
772.marian_r 100.00% 102.02% 102.60% 100.00% 100.57%
782.lbm_r 100.00% 95.14% 94.62% 100.00% 99.45%
Smaller binary sizes are always better.

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

Run-times

Benchmark Generic x86-64-v3 Native x86-64-v3 Native
709.cactus_r 100.00% 93.12% 93.98% 100.00% 100.92%
722.palm_r 100.00% 93.95% 92.90% 100.00% 98.89%
731.astcenc_r 100.00% 88.31% 92.64% 100.00% 104.90%
736.ocio_r 100.00% 79.70% 78.63% 100.00% 98.66%
737.gmsh_r 100.00% 99.33% 100.00% 100.00% 100.67%
748.flightdm_r 100.00% 98.80% 98.32% 100.00% 99.51%
749.fotonik3d_r 100.00% 86.32% 86.77% 100.00% 100.52%
765.roms_r 100.00% 79.21% 79.21% 100.00% 100.00%
766.femflow_r 100.00% 60.77% 58.46% 100.00% 96.20%
767.nest_r 100.00% 99.34% 99.34% 100.00% 100.00%
772.marian_r 100.00% 100.00% 100.00% 100.00% 100.00%
782.lbm_r 100.00% 97.79% 105.97% 100.00% 108.37%
The compiler used was GCC 17.0.0 20261015 (experimental) (revision 84348add236100e5f69a3e9ec4f7b1038bf2b83d).
Naturally, the smaller the run-times, the better.
See also continuous gcc tuning benchmarking at https://lnt.opensuse.org/db_default/v4/SPEC/spec_report/tuning.

Non-debug executable sizes

Benchmark Generic x86-64-v3 Native x86-64-v3 Native
709.cactus_r 100.00% 97.71% 98.31% 100.00% 100.62%
722.palm_r 100.00% 103.18% 103.33% 100.00% 100.15%
731.astcenc_r 100.00% 100.81% 101.10% 100.00% 100.29%
736.ocio_r 100.00% 102.45% 103.23% 100.00% 100.75%
737.gmsh_r 100.00% 101.94% 102.69% 100.00% 100.73%
748.flightdm_r 100.00% 99.81% 100.70% 100.00% 100.89%
749.fotonik3d_r 100.00% 116.74% 117.43% 100.00% 100.59%
765.roms_r 100.00% 104.39% 103.97% 100.00% 99.59%
766.femflow_r 100.00% 94.35% 94.88% 100.00% 100.57%
767.nest_r 100.00% 101.18% 101.90% 100.00% 100.71%
772.marian_r 100.00% 102.02% 102.32% 100.00% 100.29%
782.lbm_r 100.00% 95.38% 95.86% 100.00% 100.50%
Smaller binary sizes are always better.