SPEC CPU result comparison

15th October 2026 - Normalized report

SPEC2017 INTrate: Trunk LTO -Ofast - previous znver's vs native

-Ofast -flto -march=znverX vs -Ofast -flto -march=native

AMD EPYC 8635P (zen5 based Sorano)

Run-times

Benchmark Znver3 Znver4 Native Znver4 Native
500.perlbench_r 100.00% 98.04% 98.69% 100.00% 100.67%
502.gcc_r 100.00% 100.87% 100.87% 100.00% 100.00%
505.mcf_r 100.00% 100.56% 101.11% 100.00% 100.55%
520.omnetpp_r 100.00% 95.57% 99.01% 100.00% 103.61%
523.xalancbmk_r 100.00% 97.23% 97.23% 100.00% 100.00%
525.x264_r 100.00% 102.90% 101.24% 100.00% 98.39%
531.deepsjeng_r 100.00% 99.15% 98.58% 100.00% 99.43%
541.leela_r 100.00% 99.52% 99.29% 100.00% 99.76%
548.exchange2_r 100.00% 98.91% 98.91% 100.00% 100.00%
557.xz_r 100.00% 98.96% 98.96% 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 Znver3 Znver4 Native Znver4 Native
500.perlbench_r 100.00% 100.21% 100.34% 100.00% 100.13%
502.gcc_r 100.00% 99.90% 100.17% 100.00% 100.27%
505.mcf_r 100.00% 104.69% 104.69% 100.00% 100.00%
520.omnetpp_r 100.00% 100.25% 100.24% 100.00% 99.99%
523.xalancbmk_r 100.00% 98.90% 98.90% 100.00% 100.00%
525.x264_r 100.00% 98.16% 98.18% 100.00% 100.02%
531.deepsjeng_r 100.00% 100.09% 100.09% 100.00% 100.00%
541.leela_r 100.00% 99.70% 99.66% 100.00% 99.95%
548.exchange2_r 100.00% 96.70% 97.06% 100.00% 100.37%
557.xz_r 100.00% 99.14% 99.32% 100.00% 100.18%
Smaller binary sizes are always better.