SPEC CPU result comparison

15th October 2026 - Normalized report

SPEC2026 INTrate: Trunk LTO -Ofast native - using 128/256/512bit vectors vs default

-Ofast -march=native -mtune=native with and without -mprefer-vector-width=128,256 and/or 512

Please note that even though the header of this comparison says -Ofast, benchmarks 710.omnetpp_r, 734.vpr_r, 735.gem5_r and 753.ns3_r were additionally compiled with -fno-finite-math-only.

AMD EPYC 8635P (zen5 based Sorano)

Run-times

Benchmark 128 256 Default 256 Default
706.stockfish_r 100.00% 84.92% 82.54% 100.00% 97.20%
707.ntest_r 100.00% 100.00% 100.00% 100.00% 100.00%
708.sqlite_r 100.00% 101.53% 102.18% 100.00% 100.64%
710.omnetpp_r 100.00% 96.37% 97.49% 100.00% 101.16%
714.cpython_r 100.00% 106.45% 106.45% 100.00% 100.00%
721.gcc_r 100.00% 100.64% 100.64% 100.00% 100.00%
723.llvm_r 100.00% 101.29% 100.65% 100.00% 99.36%
727.cppcheck_r 100.00% 102.23% 102.23% 100.00% 100.00%
729.abc_r 100.00% 100.00% 100.00% 100.00% 100.00%
734.vpr_r 100.00% 101.04% 101.04% 100.00% 100.00%
735.gem5_r 100.00% 100.00% 101.26% 100.00% 101.26%
750.sealcrypto_r 100.00% 86.53% 80.83% 100.00% 93.41%
753.ns3_r 100.00% 101.22% 101.22% 100.00% 100.00%
777.zstd_r 100.00% 100.70% 100.70% 100.00% 100.00%
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 128 256 Default 256 Default
706.stockfish_r 100.00% 100.25% 100.48% 100.00% 100.23%
707.ntest_r 100.00% 99.94% 100.84% 100.00% 100.91%
708.sqlite_r 100.00% 101.19% 102.26% 100.00% 101.05%
710.omnetpp_r 100.00% 100.15% 100.64% 100.00% 100.50%
714.cpython_r 100.00% 99.66% 101.24% 100.00% 101.59%
721.gcc_r 100.00% 101.00% 102.52% 100.00% 101.50%
723.llvm_r 100.00% 102.39% 105.01% 100.00% 102.56%
727.cppcheck_r 100.00% 99.34% 99.53% 100.00% 100.19%
729.abc_r 100.00% 104.45% 106.21% 100.00% 101.69%
734.vpr_r 100.00% 101.44% 102.05% 100.00% 100.60%
735.gem5_r 100.00% 98.86% 99.40% 100.00% 100.55%
750.sealcrypto_r 100.00% 101.60% 106.10% 100.00% 104.43%
753.ns3_r 100.00% 99.85% 100.29% 100.00% 100.44%
777.zstd_r 100.00% 101.72% 107.35% 100.00% 105.54%
Smaller binary sizes are always better.

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

Run-times

Benchmark 128 Default
706.stockfish_r 100.00% 96.10%
707.ntest_r 100.00% 100.72%
708.sqlite_r 100.00% 100.46%
710.omnetpp_r 100.00% 101.88%
714.cpython_r 100.00% 103.26%
721.gcc_r 100.00% 100.43%
723.llvm_r 100.00% 100.63%
727.cppcheck_r 100.00% 96.88%
729.abc_r 100.00% 100.70%
734.vpr_r 100.00% 100.99%
735.gem5_r 100.00% 100.75%
750.sealcrypto_r 100.00% 85.57%
753.ns3_r 100.00% 102.09%
777.zstd_r 100.00% 101.32%
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 128 Default
706.stockfish_r 100.00% 100.17%
707.ntest_r 100.00% 99.68%
708.sqlite_r 100.00% 104.26%
710.omnetpp_r 100.00% 100.21%
714.cpython_r 100.00% 100.46%
721.gcc_r 100.00% 100.65%
723.llvm_r 100.00% 101.83%
727.cppcheck_r 100.00% 99.46%
729.abc_r 100.00% 105.52%
734.vpr_r 100.00% 101.21%
735.gem5_r 100.00% 99.09%
750.sealcrypto_r 100.00% 100.93%
753.ns3_r 100.00% 100.00%
777.zstd_r 100.00% 102.79%
Smaller binary sizes are always better.