sklbench hardware comparison dashboard

vanilla sklearn

Python 3.12.13

Packages

  • scikit-learn 1.9.0 (pypi)
  • numpy 2.4.6 (conda-forge)
  • scipy 1.17.1 (conda-forge)
  • pandas 2.3.3 (conda-forge)

Threadpools

  • libopenblas 0.3.33 (16 threads)
  • libgomp (16 threads)

OpenMP

  • GNU libgomp (OpenMP 4.5)
  • GOMP_SPINCOUNT: 300000

Full environment

view pixi env JSON

Hardware

CPU

Intel(R) Core(TM) Ultra X7 358H

X86_64, 16 physical cores, 16 logical CPUs

31 GB RAM

GPU(s)

  • level_zero:gpu:0: Intel(R) Arc(TM) B390 GPU (29 GB)

sklearnex-gpu-B390

Python 3.12.13

Packages

  • scikit-learn 1.8.0 (conda-forge)
  • scikit_learn_intelex 2026.1.0 (pypi)
  • scikit_learn_intelex_gpu 2026.1.0 (pypi)
  • dpnp 0.20.0 (pypi)
  • daal 2026.1.0 (pypi)

Threadpools

  • libopenblas 0.3.33 (16 threads)
  • libgomp (16 threads)

OpenMP

  • GNU libgomp (OpenMP 4.5)
  • GOMP_SPINCOUNT: 1

Full environment

view pixi env JSON

Hardware

CPU

Intel(R) Core(TM) Ultra X7 358H

X86_64, 16 physical cores, 16 logical CPUs

31 GB RAM

GPU(s)

  • level_zero:gpu:0: Intel(R) Arc(TM) B390 GPU (29 GB)

sklearn-dpnp-gpu-B390

Python 3.12.13

Array API active

Packages

  • scikit-learn 1.9.0 (pypi)
  • dpnp 0.20.0 (pypi)

Threadpools

  • libopenblas 0.3.33 (16 threads)
  • libgomp (16 threads)

OpenMP

  • GNU libgomp (OpenMP 4.5)
  • GOMP_SPINCOUNT: 300000

Full environment

view pixi env JSON

Hardware

CPU

Intel(R) Core(TM) Ultra X7 358H

X86_64, 16 physical cores, 16 logical CPUs

31 GB RAM

GPU(s)

  • level_zero:gpu:0: Intel(R) Arc(TM) B390 GPU (29 GB)

sklearn-torch-xpu-B390

Python 3.12.13

Array API active

Packages

  • scikit-learn 1.9.0 (pypi)
  • torch 2.13.0+xpu (pypi)

Threadpools

  • libopenblas 0.3.33 (16 threads)
  • libgomp (16 threads)

OpenMP

  • GNU libgomp (OpenMP 4.5)
  • GOMP_SPINCOUNT: 300000

Full environment

view pixi env JSON

Hardware

CPU

Intel(R) Core(TM) Ultra X7 358H

X86_64, 16 physical cores, 16 logical CPUs

31 GB RAM

GPU(s)

  • level_zero:gpu:0: Intel(R) Arc(TM) B390 GPU (29 GB)

fit / speed-up vs vanilla sklearn (62 points)

● Metrics and benchmark setup match the baseline
■ Metrics match the baseline, but some comparison details are worth reporting:
  • 🔁 Number of iteration differs: this might mean algorithms differ - LogisticRegression (3)
  • ↩ Some operations fell back to CPU according to benchmark logs - Ridge (10)
● Ran via sklearnex but fell back to stock scikit-learn, no oneDAL acceleration (3 points)
✕ Benchmark run failed, shown at the bottom of its column (6 points)

predict / speed-up vs vanilla sklearn (62 points)

● Metrics and benchmark setup match the baseline
■ Metrics match the baseline, but some comparison details are worth reporting:
  • 🔁 Number of iteration differs: this might mean algorithms differ - LogisticRegression (3)
  • ↩ Some operations fell back to CPU according to benchmark logs - Ridge (10)
● Ran via sklearnex but fell back to stock scikit-learn, no oneDAL acceleration (3 points)
✕ Benchmark run failed, shown at the bottom of its column (6 points)
Detailed results

vanilla sklearn

Python 3.12.13

Packages

  • scikit-learn 1.9.0 (pypi)
  • numpy 2.4.6 (conda-forge)
  • scipy 1.17.1 (conda-forge)
  • pandas 2.3.3 (conda-forge)

Threadpools

  • libopenblas 0.3.33 (16 threads)
  • libgomp (16 threads)

OpenMP

  • GNU libgomp (OpenMP 4.5)
  • GOMP_SPINCOUNT: 300000

Full environment

view pixi env JSON

Hardware

CPU

Intel(R) Core(TM) Ultra X7 358H

X86_64, 16 physical cores, 16 logical CPUs

31 GB RAM

GPU(s)

  • level_zero:gpu:0: Intel(R) Arc(TM) B390 GPU (29 GB)

sklearnex-cpu-laptop

Python 3.12.13

Packages

  • scikit-learn 1.8.0 (conda-forge)
  • scikit_learn_intelex 2026.1.0 (pypi)
  • daal 2026.1.0 (pypi)

Threadpools

  • libopenblas 0.3.33 (16 threads)
  • libgomp (16 threads)

OpenMP

  • GNU libgomp (OpenMP 4.5)
  • GOMP_SPINCOUNT: 1

Full environment

view pixi env JSON

Hardware

CPU

Intel(R) Core(TM) Ultra X7 358H

X86_64, 16 physical cores, 16 logical CPUs

31 GB RAM

GPU(s)

  • level_zero:gpu:0: Intel(R) Arc(TM) B390 GPU (29 GB)

sklearn-GNR

Python 3.12.13

Packages

  • scikit-learn 1.9.0 (pypi)
  • numpy 2.4.6 (conda-forge)
  • scipy 1.17.1 (conda-forge)
  • pandas 2.3.3 (conda-forge)

Threadpools

  • libopenblas 0.3.33 (128 threads)
  • libgomp (344 threads)

OpenMP

  • GNU libgomp (OpenMP 4.5)
  • GOMP_SPINCOUNT: 300000

Full environment

view pixi env JSON

Hardware

CPU

Intel(R) Xeon(R) 6787P

X86_64, 172 physical cores, 344 logical CPUs

499 GB RAM

GPU(s)

No GPU detected.

sklearnex-cpu-GNR

Python 3.12.13

Packages

  • scikit-learn 1.8.0 (conda-forge)
  • scikit_learn_intelex 2026.1.0 (pypi)
  • daal 2026.1.0 (pypi)

Threadpools

  • libopenblas 0.3.33 (128 threads)
  • libgomp (344 threads)

OpenMP

  • GNU libgomp (OpenMP 4.5)
  • GOMP_SPINCOUNT: 300000

Full environment

view pixi env JSON

Hardware

CPU

Intel(R) Xeon(R) 6787P

X86_64, 172 physical cores, 344 logical CPUs

499 GB RAM

GPU(s)

No GPU detected.

linear / fit (122 points)

● Metrics and benchmark setup match the baseline
■ Metrics match the baseline, but some comparison details are worth reporting:
  • 🔁 Number of iteration differs: this might mean algorithms differ - LogisticRegression (21)
◇ Metrics differ from the baseline (5 points)
● Ran via sklearnex but fell back to stock scikit-learn, no oneDAL acceleration (20 points)

linear / predict (122 points)

● Metrics and benchmark setup match the baseline
■ Metrics match the baseline, but some comparison details are worth reporting:
  • 🔁 Number of iteration differs: this might mean algorithms differ - LogisticRegression (21)
● Ran via sklearnex but fell back to stock scikit-learn, no oneDAL acceleration (20 points)
Detailed results

tree-based / fit (80 points)

● Metrics and benchmark setup match the baseline
■ Metrics match the baseline, but some comparison details are worth reporting:
  • 🧺 Scikit-learn intelex uses binning & histogram-based splits while scikit-learn doesn't - RandomForestClassifier (10), ExtraTreesRegressor (9), ExtraTreesClassifier (8), RandomForestRegressor (4)
◇ Metrics differ from the baseline (39 points)
● Ran via sklearnex but fell back to stock scikit-learn, no oneDAL acceleration (3 points)
✕ Benchmark run failed, shown at the bottom of its column (4 points)

tree-based / predict (80 points)

● Metrics and benchmark setup match the baseline
■ Metrics match the baseline, but some comparison details are worth reporting:
  • 🧺 Scikit-learn intelex uses binning & histogram-based splits while scikit-learn doesn't - RandomForestClassifier (10), ExtraTreesRegressor (9), ExtraTreesClassifier (8), RandomForestRegressor (4)
● Ran via sklearnex but fell back to stock scikit-learn, no oneDAL acceleration (3 points)
✕ Benchmark run failed, shown at the bottom of its column (4 points)
Detailed results

clustering / fit (29 points)

● Metrics and benchmark setup match the baseline
■ Metrics match the baseline, but some comparison details are worth reporting:
  • 🔁 Number of iteration differs: this might mean algorithms differ - KMeans (8)
✕ Benchmark run failed, shown at the bottom of its column (6 points)

clustering / predict (29 points)

● Metrics and benchmark setup match the baseline
■ Metrics match the baseline, but some comparison details are worth reporting:
  • 🔁 Number of iteration differs: this might mean algorithms differ - KMeans (8)
✕ Benchmark run failed, shown at the bottom of its column (6 points)
Detailed results