AMD EPYC 4th Gen AVX-512 Comparison

AMD EPYC 9654 Genoa AVX-512 benchmark comparison by Michael Larabel for a future article.

HTML result view exported from: https://openbenchmarking.org/result/2212195-NE-AVXCOMPAR69&grs&sro.

TensorFlow

Device: CPU - Batch Size: 16 - Model: AlexNet

oneDNN

Harness: Recurrent Neural Network Training - Data Type: bf16bf16bf16 - Engine: CPU

AI Benchmark Alpha

Device Training Score

oneDNN

Harness: Recurrent Neural Network Training - Data Type: f32 - Engine: CPU

OpenVINO

Model: Weld Porosity Detection FP16 - Device: CPU

OpenVINO

Model: Weld Porosity Detection FP16 - Device: CPU

OpenVINO

Model: Face Detection FP16 - Device: CPU

OpenVINO

Model: Face Detection FP16 - Device: CPU

OpenVINO

Model: Age Gender Recognition Retail 0013 FP16 - Device: CPU

Cpuminer-Opt

Algorithm: LBC, LBRY Credits

AI Benchmark Alpha

Device AI Score

OpenVINO

Model: Machine Translation EN To DE FP16 - Device: CPU

OpenVINO

Model: Machine Translation EN To DE FP16 - Device: CPU

OpenVINO

Model: Weld Porosity Detection FP16-INT8 - Device: CPU

OpenVINO

Model: Weld Porosity Detection FP16-INT8 - Device: CPU

OpenVINO

Model: Face Detection FP16-INT8 - Device: CPU

OpenVINO

Model: Face Detection FP16-INT8 - Device: CPU

OpenVINO

Model: Vehicle Detection FP16 - Device: CPU

OpenVINO

Model: Vehicle Detection FP16 - Device: CPU

AI Benchmark Alpha

Device Inference Score

OpenVINO

Model: Vehicle Detection FP16-INT8 - Device: CPU

OpenVINO

Model: Vehicle Detection FP16-INT8 - Device: CPU

oneDNN

Harness: Deconvolution Batch shapes_3d - Data Type: f32 - Engine: CPU

OpenVINO

Model: Person Vehicle Bike Detection FP16 - Device: CPU

OpenVINO

Model: Person Vehicle Bike Detection FP16 - Device: CPU

TensorFlow

Device: CPU - Batch Size: 16 - Model: ResNet-50

OSPRay

Benchmark: gravity_spheres_volume/dim_512/scivis/real_time

Cpuminer-Opt

Algorithm: Quad SHA-256, Pyrite

OpenVINO

Model: Person Detection FP16 - Device: CPU

OpenVINO

Model: Person Detection FP16 - Device: CPU

OpenVINO

Model: Person Detection FP32 - Device: CPU

OpenVINO

Model: Person Detection FP32 - Device: CPU

OSPRay

Benchmark: gravity_spheres_volume/dim_512/ao/real_time

Cpuminer-Opt

Algorithm: scrypt

OpenVINO

Model: Age Gender Recognition Retail 0013 FP16-INT8 - Device: CPU

Mobile Neural Network

Model: resnet-v2-50

Neural Magic DeepSparse

Model: NLP Question Answering, BERT base uncased SQuaD 12layer Pruned90 - Scenario: Asynchronous Multi-Stream

Neural Magic DeepSparse

Model: NLP Question Answering, BERT base uncased SQuaD 12layer Pruned90 - Scenario: Asynchronous Multi-Stream

Mobile Neural Network

Model: inception-v3

Cpuminer-Opt

Algorithm: Garlicoin

miniBUDE

Implementation: OpenMP - Input Deck: BM1

miniBUDE

Implementation: OpenMP - Input Deck: BM1

Neural Magic DeepSparse

Model: NLP Question Answering, BERT base uncased SQuaD 12layer Pruned90 - Scenario: Synchronous Single-Stream

Neural Magic DeepSparse

Model: NLP Question Answering, BERT base uncased SQuaD 12layer Pruned90 - Scenario: Synchronous Single-Stream

Neural Magic DeepSparse

Model: CV Classification, ResNet-50 ImageNet - Scenario: Asynchronous Multi-Stream

Neural Magic DeepSparse

Model: CV Classification, ResNet-50 ImageNet - Scenario: Asynchronous Multi-Stream

oneDNN

Harness: Recurrent Neural Network Inference - Data Type: bf16bf16bf16 - Engine: CPU

OpenVINO

Model: Age Gender Recognition Retail 0013 FP16 - Device: CPU

Cpuminer-Opt

Algorithm: Skeincoin

miniBUDE

Implementation: OpenMP - Input Deck: BM2

miniBUDE

Implementation: OpenMP - Input Deck: BM2

OSPRay

Benchmark: gravity_spheres_volume/dim_512/pathtracer/real_time

Cpuminer-Opt

Algorithm: x25x

simdjson

Throughput Test: DistinctUserID

simdjson

Throughput Test: PartialTweets

Neural Magic DeepSparse

Model: CV Detection,YOLOv5s COCO - Scenario: Asynchronous Multi-Stream

Neural Magic DeepSparse

Model: CV Detection,YOLOv5s COCO - Scenario: Asynchronous Multi-Stream

simdjson

Throughput Test: Kostya

OSPRay Studio

Camera: 2 - Resolution: 1080p - Samples Per Pixel: 32 - Renderer: Path Tracer

OSPRay Studio

Camera: 2 - Resolution: 4K - Samples Per Pixel: 16 - Renderer: Path Tracer

OSPRay Studio

Camera: 2 - Resolution: 1080p - Samples Per Pixel: 16 - Renderer: Path Tracer

OSPRay Studio

Camera: 1 - Resolution: 1080p - Samples Per Pixel: 32 - Renderer: Path Tracer

OSPRay Studio

Camera: 1 - Resolution: 1080p - Samples Per Pixel: 1 - Renderer: Path Tracer

OSPRay Studio

Camera: 2 - Resolution: 4K - Samples Per Pixel: 32 - Renderer: Path Tracer

OSPRay Studio

Camera: 1 - Resolution: 4K - Samples Per Pixel: 1 - Renderer: Path Tracer

OSPRay Studio

Camera: 1 - Resolution: 4K - Samples Per Pixel: 32 - Renderer: Path Tracer

OSPRay Studio

Camera: 1 - Resolution: 4K - Samples Per Pixel: 16 - Renderer: Path Tracer

OSPRay Studio

Camera: 1 - Resolution: 1080p - Samples Per Pixel: 16 - Renderer: Path Tracer

OSPRay Studio

Camera: 3 - Resolution: 1080p - Samples Per Pixel: 16 - Renderer: Path Tracer

OSPRay Studio

Camera: 3 - Resolution: 1080p - Samples Per Pixel: 32 - Renderer: Path Tracer

simdjson

Throughput Test: TopTweet

OSPRay Studio

Camera: 3 - Resolution: 4K - Samples Per Pixel: 1 - Renderer: Path Tracer

simdjson

Throughput Test: LargeRandom

Neural Magic DeepSparse

Model: NLP Text Classification, BERT base uncased SST2 - Scenario: Asynchronous Multi-Stream

OSPRay Studio

Camera: 3 - Resolution: 4K - Samples Per Pixel: 16 - Renderer: Path Tracer

Neural Magic DeepSparse

Model: NLP Text Classification, BERT base uncased SST2 - Scenario: Asynchronous Multi-Stream

OSPRay Studio

Camera: 3 - Resolution: 1080p - Samples Per Pixel: 1 - Renderer: Path Tracer

OSPRay Studio

Camera: 3 - Resolution: 4K - Samples Per Pixel: 32 - Renderer: Path Tracer

Embree

Binary: Pathtracer ISPC - Model: Asian Dragon

Embree

Binary: Pathtracer ISPC - Model: Asian Dragon Obj

Neural Magic DeepSparse

Model: CV Detection,YOLOv5s COCO - Scenario: Synchronous Single-Stream

Neural Magic DeepSparse

Model: CV Detection,YOLOv5s COCO - Scenario: Synchronous Single-Stream

Embree

Binary: Pathtracer ISPC - Model: Crown

OSPRay Studio

Camera: 2 - Resolution: 4K - Samples Per Pixel: 1 - Renderer: Path Tracer

OSPRay Studio

Camera: 2 - Resolution: 1080p - Samples Per Pixel: 1 - Renderer: Path Tracer

Neural Magic DeepSparse

Model: NLP Text Classification, DistilBERT mnli - Scenario: Asynchronous Multi-Stream

OpenFOAM

Input: drivaerFastback, Medium Mesh Size - Execution Time

Neural Magic DeepSparse

Model: NLP Text Classification, DistilBERT mnli - Scenario: Asynchronous Multi-Stream

Neural Magic DeepSparse

Model: NLP Text Classification, BERT base uncased SST2 - Scenario: Synchronous Single-Stream

Neural Magic DeepSparse

Model: NLP Text Classification, BERT base uncased SST2 - Scenario: Synchronous Single-Stream

OpenVKL

Benchmark: vklBenchmark ISPC

OpenVINO

Model: Age Gender Recognition Retail 0013 FP16-INT8 - Device: CPU

Neural Magic DeepSparse

Model: NLP Token Classification, BERT base uncased conll2003 - Scenario: Synchronous Single-Stream

Neural Magic DeepSparse

Model: NLP Token Classification, BERT base uncased conll2003 - Scenario: Synchronous Single-Stream

Neural Magic DeepSparse

Model: NLP Document Classification, oBERT base uncased on IMDB - Scenario: Synchronous Single-Stream

Neural Magic DeepSparse

Model: NLP Document Classification, oBERT base uncased on IMDB - Scenario: Synchronous Single-Stream

NCNN

Target: CPU - Model: blazeface

LeelaChessZero

Backend: Eigen

NCNN

Target: CPU - Model: regnety_400m

SMHasher

Hash: FarmHash32 x86_64 AVX

NCNN

Target: CPU - Model: efficientnet-b0

ONNX Runtime

Model: fcn-resnet101-11 - Device: CPU - Executor: Standard

LeelaChessZero

Backend: BLAS

CP2K Molecular Dynamics

Input: Fayalite-FIST

Mobile Neural Network

Model: SqueezeNetV1.0

OpenFOAM

Input: drivaerFastback, Medium Mesh Size - Mesh Time

Neural Magic DeepSparse

Model: NLP Text Classification, DistilBERT mnli - Scenario: Synchronous Single-Stream

Neural Magic DeepSparse

Model: NLP Text Classification, DistilBERT mnli - Scenario: Synchronous Single-Stream

Numpy Benchmark

JPEG XL libjxl

Input: JPEG - Quality: 90

NCNN

Target: CPU - Model: googlenet

JPEG XL libjxl

Input: PNG - Quality: 90

NCNN

Target: CPU - Model: mnasnet

Numenta Anomaly Benchmark

Detector: Bayesian Changepoint

oneDNN

Harness: Convolution Batch Shapes Auto - Data Type: f32 - Engine: CPU

Intel Open Image Denoise

Run: RTLightmap.hdr.4096x4096

ONNX Runtime

Model: bertsquad-12 - Device: CPU - Executor: Standard

Numenta Anomaly Benchmark

Detector: Windowed Gaussian

JPEG XL libjxl

Input: JPEG - Quality: 100

Darmstadt Automotive Parallel Heterogeneous Suite

Backend: OpenMP - Kernel: NDT Mapping

NCNN

Target: CPU - Model: resnet50

GROMACS

Implementation: MPI CPU - Input: water_GMX50_bare

Numenta Anomaly Benchmark

Detector: Relative Entropy

oneDNN

Harness: Deconvolution Batch shapes_1d - Data Type: f32 - Engine: CPU

Intel Open Image Denoise

Run: RT.hdr_alb_nrm.3840x2160

Intel Open Image Denoise

Run: RT.ldr_alb_nrm.3840x2160

CPU Temperature Monitor

Phoronix Test Suite System Monitoring

CPU Power Consumption Monitor

Phoronix Test Suite System Monitoring