# AMD Threadripper PRO 5975WX (Zen 3), Ubuntu 24.04, 8 CPUs (taskset -c 32-39), Python 3.12.3, scikit-learn 1.9.1, numpy 2.5.4, pandas 3.0.6, joblib 1.6.0, 12 October 2026
# lightgbm-takt 4.7.0 (free) from PyPI, same model and data as zen3-official.txt

$ python pi_cost.py all work 3 ...
N = 8 CPUs, 3 rounds per setting, median wall time of the permutation_importance call
numpy, n_jobs=1, 1 thread                           22.69 s   5f242b6b5bd26ee0 same result
pandas, n_jobs=1, 1 thread                          22.83 s   5f242b6b5bd26ee0 same result
numpy, n_jobs=N, 1 thread                            7.84 s   5f242b6b5bd26ee0 same result
pandas, n_jobs=N, 1 thread                           7.92 s   5f242b6b5bd26ee0 same result
pandas, n_jobs=1, LightGBM default threads           9.24 s   5f242b6b5bd26ee0 same result
numpy, n_jobs=1, N threads                           9.44 s   5f242b6b5bd26ee0 same result

# alternating runs, official wheel (off) and lightgbm-takt (lgbt): seconds, result hash, largest mean importance
# setting "numpy, n_jobs=1, N threads"
off 8.655 5f242b6b5bd26ee0 0.025123299949183765
lgbt 9.287 5f242b6b5bd26ee0 0.025123299949183765
off 8.700 5f242b6b5bd26ee0 0.025123299949183765
lgbt 9.542 5f242b6b5bd26ee0 0.025123299949183765
off 8.658 5f242b6b5bd26ee0 0.025123299949183765
lgbt 9.490 5f242b6b5bd26ee0 0.025123299949183765
off 8.565 5f242b6b5bd26ee0 0.025123299949183765
lgbt 9.302 5f242b6b5bd26ee0 0.025123299949183765
off 8.746 5f242b6b5bd26ee0 0.025123299949183765
lgbt 9.392 5f242b6b5bd26ee0 0.025123299949183765
off 8.633 5f242b6b5bd26ee0 0.025123299949183765
lgbt 9.284 5f242b6b5bd26ee0 0.025123299949183765
# setting "pandas, n_jobs=1, LightGBM default threads"
off 9.050 5f242b6b5bd26ee0 0.025123299949183765
lgbt 9.273 5f242b6b5bd26ee0 0.025123299949183765
off 9.058 5f242b6b5bd26ee0 0.025123299949183765
lgbt 9.276 5f242b6b5bd26ee0 0.025123299949183765
off 8.966 5f242b6b5bd26ee0 0.025123299949183765
lgbt 9.241 5f242b6b5bd26ee0 0.025123299949183765
off 9.000 5f242b6b5bd26ee0 0.025123299949183765
lgbt 9.232 5f242b6b5bd26ee0 0.025123299949183765
off 9.009 5f242b6b5bd26ee0 0.025123299949183765
lgbt 9.300 5f242b6b5bd26ee0 0.025123299949183765
off 8.998 5f242b6b5bd26ee0 0.025123299949183765
lgbt 9.323 5f242b6b5bd26ee0 0.025123299949183765