SPEC CPU result comparison

15th October 2026 - Normalized report

SPEC2026 INTrate: 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
706.stockfish_r 100.00% 76.40% 64.60% 100.00% 84.55%
707.ntest_r 100.00% 89.01% 88.37% 100.00% 99.29%
708.sqlite_r 100.00% 101.54% 102.42% 100.00% 100.87%
710.omnetpp_r 100.00% 97.51% 97.51% 100.00% 100.00%
714.cpython_r 100.00% 103.92% 105.15% 100.00% 101.18%
721.gcc_r 100.00% 100.00% 100.00% 100.00% 100.00%
723.llvm_r 100.00% 101.29% 101.94% 100.00% 100.64%
727.cppcheck_r 100.00% 98.91% 97.82% 100.00% 98.90%
729.abc_r 100.00% 99.09% 99.09% 100.00% 100.00%
734.vpr_r 100.00% 96.73% 97.73% 100.00% 101.04%
735.gem5_r 100.00% 100.00% 100.00% 100.00% 100.00%
750.sealcrypto_r 100.00% 93.24% 75.36% 100.00% 80.83%
753.ns3_r 100.00% 100.00% 100.00% 100.00% 100.00%
777.zstd_r 100.00% 102.17% 103.62% 100.00% 101.42%
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
706.stockfish_r 100.00% 100.40% 99.25% 100.00% 98.86%
707.ntest_r 100.00% 99.39% 100.85% 100.00% 101.47%
708.sqlite_r 100.00% 106.20% 106.39% 100.00% 100.18%
710.omnetpp_r 100.00% 101.92% 103.04% 100.00% 101.10%
714.cpython_r 100.00% 105.55% 107.69% 100.00% 102.03%
721.gcc_r 100.00% 102.09% 104.42% 100.00% 102.28%
723.llvm_r 100.00% 106.14% 109.70% 100.00% 103.35%
727.cppcheck_r 100.00% 100.51% 101.69% 100.00% 101.17%
729.abc_r 100.00% 111.50% 112.35% 100.00% 100.76%
734.vpr_r 100.00% 103.88% 104.98% 100.00% 101.06%
735.gem5_r 100.00% 101.47% 102.76% 100.00% 101.28%
750.sealcrypto_r 100.00% 106.63% 112.40% 100.00% 105.41%
753.ns3_r 100.00% 101.43% 101.49% 100.00% 100.06%
777.zstd_r 100.00% 114.75% 117.19% 100.00% 102.13%
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
706.stockfish_r 100.00% 80.65% 90.32% 100.00% 112.00%
707.ntest_r 100.00% 88.31% 87.68% 100.00% 99.29%
708.sqlite_r 100.00% 100.46% 101.62% 100.00% 101.15%
710.omnetpp_r 100.00% 100.00% 98.16% 100.00% 98.16%
714.cpython_r 100.00% 106.08% 104.30% 100.00% 98.33%
721.gcc_r 100.00% 100.43% 100.00% 100.00% 99.57%
723.llvm_r 100.00% 100.63% 101.27% 100.00% 100.63%
727.cppcheck_r 100.00% 98.53% 100.84% 100.00% 102.35%
729.abc_r 100.00% 100.00% 100.00% 100.00% 100.00%
734.vpr_r 100.00% 99.04% 98.32% 100.00% 99.27%
735.gem5_r 100.00% 100.00% 98.49% 100.00% 98.49%
750.sealcrypto_r 100.00% 88.78% 87.76% 100.00% 98.85%
753.ns3_r 100.00% 99.49% 100.00% 100.00% 100.51%
777.zstd_r 100.00% 98.06% 98.92% 100.00% 100.88%
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
706.stockfish_r 100.00% 100.40% 99.63% 100.00% 99.23%
707.ntest_r 100.00% 99.39% 99.48% 100.00% 100.09%
708.sqlite_r 100.00% 106.19% 109.02% 100.00% 102.66%
710.omnetpp_r 100.00% 101.92% 102.96% 100.00% 101.02%
714.cpython_r 100.00% 105.55% 106.29% 100.00% 100.70%
721.gcc_r 100.00% 102.09% 103.07% 100.00% 100.96%
723.llvm_r 100.00% 106.14% 106.96% 100.00% 100.77%
727.cppcheck_r 100.00% 100.51% 101.49% 100.00% 100.97%
729.abc_r 100.00% 111.50% 111.75% 100.00% 100.22%
734.vpr_r 100.00% 103.88% 104.48% 100.00% 100.57%
735.gem5_r 100.00% 101.47% 102.24% 100.00% 100.76%
750.sealcrypto_r 100.00% 106.63% 107.34% 100.00% 100.67%
753.ns3_r 100.00% 101.43% 101.13% 100.00% 99.71%
777.zstd_r 100.00% 114.73% 114.98% 100.00% 100.21%
Smaller binary sizes are always better.