SPEC CPU result comparison

15th October 2026 - Normalized report

SPEC2017 INTrate: Trunk LTO -O2 - generic vs x86-64-v3 vs native

-O2 -flto vs -O2 -flto and a given -march

AMD EPYC 8635P (zen5 based Sorano)

Run-times

Benchmark Generic x86-64-v3 Native x86-64-v3 Native
500.perlbench_r 100.00% 100.67% 99.33% 100.00% 98.68%
502.gcc_r 100.00% 100.00% 100.00% 100.00% 100.00%
505.mcf_r 100.00% 100.00% 99.52% 100.00% 99.52%
520.omnetpp_r 100.00% 100.86% 100.00% 100.00% 99.15%
523.xalancbmk_r 100.00% 100.00% 103.33% 100.00% 103.33%
525.x264_r 100.00% 94.57% 95.25% 100.00% 100.72%
531.deepsjeng_r 100.00% 94.39% 93.85% 100.00% 99.43%
541.leela_r 100.00% 100.45% 100.45% 100.00% 100.00%
548.exchange2_r 100.00% 96.02% 97.57% 100.00% 101.61%
557.xz_r 100.00% 99.16% 99.16% 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 Generic x86-64-v3 Native x86-64-v3 Native
500.perlbench_r 100.00% 99.89% 100.64% 100.00% 100.75%
502.gcc_r 100.00% 100.17% 101.04% 100.00% 100.86%
505.mcf_r 100.00% 100.27% 102.24% 100.00% 101.96%
520.omnetpp_r 100.00% 100.70% 101.70% 100.00% 100.99%
523.xalancbmk_r 100.00% 100.47% 101.87% 100.00% 101.39%
525.x264_r 100.00% 98.78% 99.14% 100.00% 100.37%
531.deepsjeng_r 100.00% 100.02% 100.03% 100.00% 100.00%
541.leela_r 100.00% 100.55% 101.77% 100.00% 101.21%
548.exchange2_r 100.00% 106.98% 107.49% 100.00% 100.47%
557.xz_r 100.00% 100.42% 104.46% 100.00% 104.02%
Smaller binary sizes are always better.

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

Run-times

Benchmark Generic x86-64-v3
500.perlbench_r 100.00% 102.83%
502.gcc_r 100.00% 99.36%
505.mcf_r 100.00% 99.52%
520.omnetpp_r 100.00% 100.84%
523.xalancbmk_r 100.00% 101.12%
525.x264_r 100.00% 96.59%
531.deepsjeng_r 100.00% 96.12%
541.leela_r 100.00% 100.64%
548.exchange2_r 100.00% 94.75%
557.xz_r 100.00% 99.58%
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
500.perlbench_r 100.00% 99.89%
502.gcc_r 100.00% 100.17%
505.mcf_r 100.00% 100.27%
520.omnetpp_r 100.00% 100.70%
523.xalancbmk_r 100.00% 100.47%
525.x264_r 100.00% 98.79%
531.deepsjeng_r 100.00% 100.02%
541.leela_r 100.00% 100.60%
548.exchange2_r 100.00% 106.98%
557.xz_r 100.00% 100.46%
Smaller binary sizes are always better.