SPEC CPU result comparison

15th October 2026 - Normalized report

SPEC2026 FPrate: LTO -Ofast native - LLVM, ICX and GCC

-Ofast -march=native -flto

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 ICX GCC ICX GCC
709.cactus_r 100.00% 37.75% 115.26% 100.00% 305.32%
722.palm_r 100.00% 93.45% 120.34% 100.00% 128.78%
731.astcenc_r 100.00% 88.55% 117.18% 100.00% 132.33%
736.ocio_r 100.00% 101.08% 94.04% 100.00% 93.03%
737.gmsh_r 100.00% 89.92% 93.28% 100.00% 103.74%
748.flightdm_r 100.00% 85.78% 92.78% 100.00% 108.16%
749.fotonik3d_r 100.00% 109.38% 102.68% 100.00% 93.87%
765.roms_r 100.00% 208.02% 93.25% 100.00% 44.83%
766.femflow_r 100.00% 85.71% 95.92% 100.00% 111.90%
767.nest_r 100.00% 92.05% 98.08% 100.00% 106.55%
772.marian_r 100.00% 85.71% 64.29% 100.00% 75.00%
782.lbm_r 100.00% 45.61% 60.23% 100.00% 132.05%
LLVM clang version 23.1.3 (/labs/mjambor/git/llvm 0d261d1ca552c95a8f007e061c787ac7132fbcbc), flang version 23.1.3 (/labs/mjambor/git/llvm 0d261d1ca552c95a8f007e061c787ac7132fbcbc)
ICX Intel(R) oneAPI DPC++/C++ Compiler 2026.1.1 (2026.1.1.20260724), ifx (IFX) 2026.1.1 20260724
GCC GCC 17.0.0 20261015 (experimental) (revision 84348add236100e5f69a3e9ec4f7b1038bf2b83d)
Naturally, the smaller the run-times, the better.

Non-debug executable sizes

Benchmark LLVM ICX GCC ICX GCC
709.cactus_r 100.00% 144.37% 95.13% 100.00% 65.89%
722.palm_r 100.00% 83.57% 82.60% 100.00% 98.84%
731.astcenc_r 100.00% 130.78% 138.64% 100.00% 106.01%
736.ocio_r 100.00% 99.38% 90.22% 100.00% 90.78%
737.gmsh_r 100.00% 97.60% 96.65% 100.00% 99.03%
748.flightdm_r 100.00% 121.06% 91.45% 100.00% 75.54%
749.fotonik3d_r 100.00% 92.13% 35.38% 100.00% 38.40%
765.roms_r 100.00% 91.26% 70.58% 100.00% 77.34%
766.femflow_r 100.00% 77.85% 67.62% 100.00% 86.86%
767.nest_r 100.00% 101.43% 84.99% 100.00% 83.80%
772.marian_r 100.00% 98.33% 93.60% 100.00% 95.19%
782.lbm_r 100.00% 109.09% 45.12% 100.00% 41.36%
Smaller binary sizes are always better.

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

Run-times

Benchmark LLVM ICX GCC ICX GCC
709.cactus_r 100.00% 37.68% 100.00% 100.00% 265.41%
722.palm_r 100.00% 97.34% 127.24% 100.00% 130.72%
731.astcenc_r 100.00% 106.75% 103.99% 100.00% 97.41%
736.ocio_r 100.00% 94.85% 92.53% 100.00% 97.55%
737.gmsh_r 100.00% 90.27% 95.14% 100.00% 105.39%
748.flightdm_r 100.00% 95.84% 94.69% 100.00% 98.80%
749.fotonik3d_r 100.00% 103.48% 104.28% 100.00% 100.78%
765.roms_r 100.00% 119.80% 92.62% 100.00% 77.31%
766.femflow_r 100.00% 100.00% 117.19% 100.00% 117.19%
767.nest_r 100.00% 101.88% 101.88% 100.00% 100.00%
772.marian_r 100.00% 104.00% 116.00% 100.00% 111.54%
782.lbm_r 100.00% 87.66% 114.56% 100.00% 130.69%
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)
ICX Intel(R) oneAPI DPC++/C++ Compiler 2026.1.1 (2026.1.1.20260724), ifx (IFX) 2026.1.1 20260724
GCC GCC 17.0.0 20261015 (experimental) (revision 84348add236100e5f69a3e9ec4f7b1038bf2b83d)
Naturally, the smaller the run-times, the better.

Non-debug executable sizes

Benchmark LLVM ICX GCC ICX GCC
709.cactus_r 100.00% 141.35% 96.35% 100.00% 68.17%
722.palm_r 100.00% 90.23% 83.41% 100.00% 92.44%
731.astcenc_r 100.00% 135.27% 160.42% 100.00% 118.59%
736.ocio_r 100.00% 99.13% 91.86% 100.00% 92.67%
737.gmsh_r 100.00% 98.51% 97.57% 100.00% 99.04%
748.flightdm_r 100.00% 121.42% 91.77% 100.00% 75.58%
749.fotonik3d_r 100.00% 96.48% 36.03% 100.00% 37.34%
765.roms_r 100.00% 99.92% 71.87% 100.00% 71.92%
766.femflow_r 100.00% 82.22% 67.11% 100.00% 81.62%
767.nest_r 100.00% 102.45% 87.55% 100.00% 85.46%
772.marian_r 100.00% 96.97% 92.87% 100.00% 95.77%
782.lbm_r 100.00% 151.52% 70.28% 100.00% 46.38%
Smaller binary sizes are always better.