epyc 9654 AMD March

2 x AMD EPYC 9654 96-Core testing with a AMD Titanite_4G (RTI1004D BIOS) and ASPEED on Ubuntu 23.04 via the Phoronix Test Suite.

HTML result view exported from: https://openbenchmarking.org/result/2303292-NE-EPYC9654A80&sro&grs.

OpenCV

Test: Core

PostgreSQL

Scaling Factor: 100 - Clients: 1000 - Mode: Read Write - Average Latency

PostgreSQL

Scaling Factor: 100 - Clients: 1000 - Mode: Read Write

PostgreSQL

Scaling Factor: 100 - Clients: 800 - Mode: Read Write

PostgreSQL

Scaling Factor: 100 - Clients: 800 - Mode: Read Write - Average Latency

OpenCV

Test: Video

OpenCV

Test: Object Detection

OpenCV

Test: Image Processing

MariaDB

Clients: 512

MariaDB

Clients: 1024

MariaDB

Clients: 2048

RocksDB

Test: Random Fill Sync

TensorFlow

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

TensorFlow

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

TensorFlow

Device: CPU - Batch Size: 32 - Model: GoogLeNet

OpenCV

Test: DNN - Deep Neural Network

Neural Magic DeepSparse

Model: NLP Token Classification, BERT base uncased conll2003 - Scenario: Asynchronous Multi-Stream

Neural Magic DeepSparse

Model: NLP Document Classification, oBERT base uncased on IMDB - Scenario: Asynchronous Multi-Stream

OpenSSL

Algorithm: ChaCha20

OpenSSL

Algorithm: RSA4096

OpenSSL

Algorithm: RSA4096

OpenSSL

Algorithm: AES-256-GCM

OpenSSL

Algorithm: SHA256

RocksDB

Test: Random Read

OpenSSL

Algorithm: AES-128-GCM

OpenSSL

Algorithm: ChaCha20-Poly1305

OpenSSL

Algorithm: SHA512

MariaDB

Clients: 4096

Neural Magic DeepSparse

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

Neural Magic DeepSparse

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

Neural Magic DeepSparse

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

John The Ripper

Test: bcrypt

John The Ripper

Test: WPA PSK

John The Ripper

Test: Blowfish

TensorFlow

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

ONNX Runtime

Model: ResNet50 v1-12-int8 - Device: CPU - Executor: Parallel

Neural Magic DeepSparse

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

OpenCV

Test: Graph API

SPECFEM3D

Model: Water-layered Halfspace

Neural Magic DeepSparse

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

Neural Magic DeepSparse

Model: CV Segmentation, 90% Pruned YOLACT Pruned - Scenario: Asynchronous Multi-Stream

TensorFlow

Device: CPU - Batch Size: 32 - Model: AlexNet

RocksDB

Test: Read While Writing

ONNX Runtime

Model: ArcFace ResNet-100 - Device: CPU - Executor: Parallel

ONNX Runtime

Model: CaffeNet 12-int8 - Device: CPU - Executor: Parallel

SPECFEM3D

Model: Mount St. Helens

Neural Magic DeepSparse

Model: NLP Sentiment Analysis, 80% Pruned Quantized BERT Base Uncased - Scenario: Asynchronous Multi-Stream

TensorFlow

Device: CPU - Batch Size: 32 - Model: ResNet-50

TensorFlow

Device: CPU - Batch Size: 64 - Model: GoogLeNet

John The Ripper

Test: MD5

SPECFEM3D

Model: Tomographic Model

SPECFEM3D

Model: Homogeneous Halfspace

GROMACS

Implementation: MPI CPU - Input: water_GMX50_bare

ONNX Runtime

Model: ResNet50 v1-12-int8 - Device: CPU - Executor: Standard

ONNX Runtime

Model: ArcFace ResNet-100 - Device: CPU - Executor: Standard

Embree

Binary: Pathtracer - Model: Crown

SPECFEM3D

Model: Layered Halfspace

OpenCV

Test: Features 2D

RocksDB

Test: Read Random Write Random

Embree

Binary: Pathtracer - Model: Asian Dragon Obj

Embree

Binary: Pathtracer ISPC - Model: Crown

Embree

Binary: Pathtracer - Model: Asian Dragon

Embree

Binary: Pathtracer ISPC - Model: Asian Dragon Obj

Embree

Binary: Pathtracer ISPC - Model: Asian Dragon

ONNX Runtime

Model: yolov4 - Device: CPU - Executor: Parallel

TensorFlow

Device: CPU - Batch Size: 64 - Model: ResNet-50

MariaDB

Clients: 8192

Memcached

Set To Get Ratio: 1:5

RocksDB

Test: Sequential Fill

Darmstadt Automotive Parallel Heterogeneous Suite

Backend: OpenMP - Kernel: Points2Image

RocksDB

Test: Random Fill

RocksDB

Test: Update Random

ONNX Runtime

Model: yolov4 - Device: CPU - Executor: Standard

Apache HTTP Server

Concurrent Requests: 500

TensorFlow

Device: CPU - Batch Size: 64 - Model: AlexNet

ONNX Runtime

Model: GPT-2 - Device: CPU - Executor: Parallel

ONNX Runtime

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

OpenCV

Test: Stitching

PostgreSQL

Scaling Factor: 1 - Clients: 1000 - Mode: Read Write - Average Latency

PostgreSQL

Scaling Factor: 1 - Clients: 1000 - Mode: Read Write

ONNX Runtime

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

PostgreSQL

Scaling Factor: 1 - Clients: 800 - Mode: Read Write

PostgreSQL

Scaling Factor: 1 - Clients: 800 - Mode: Read Write - Average Latency

ONNX Runtime

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

TensorFlow

Device: CPU - Batch Size: 512 - Model: AlexNet

ONNX Runtime

Model: CaffeNet 12-int8 - Device: CPU - Executor: Standard

Timed LLVM Compilation

Build System: Ninja

Timed Node.js Compilation

Time To Compile

ONNX Runtime

Model: super-resolution-10 - Device: CPU - Executor: Parallel

Darmstadt Automotive Parallel Heterogeneous Suite

Backend: OpenMP - Kernel: NDT Mapping

ONNX Runtime

Model: GPT-2 - Device: CPU - Executor: Standard

ONNX Runtime

Model: Faster R-CNN R-50-FPN-int8 - Device: CPU - Executor: Parallel

nginx

Connections: 500

TensorFlow

Device: CPU - Batch Size: 256 - Model: GoogLeNet

Timed FFmpeg Compilation

Time To Compile

PostgreSQL

Scaling Factor: 100 - Clients: 800 - Mode: Read Only

PostgreSQL

Scaling Factor: 100 - Clients: 800 - Mode: Read Only - Average Latency

Apache HTTP Server

Concurrent Requests: 200

ONNX Runtime

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

ClickHouse

100M Rows Hits Dataset, Third Run

ClickHouse

100M Rows Hits Dataset, Second Run

PostgreSQL

Scaling Factor: 100 - Clients: 1000 - Mode: Read Only - Average Latency

PostgreSQL

Scaling Factor: 100 - Clients: 1000 - Mode: Read Only

TensorFlow

Device: CPU - Batch Size: 256 - Model: ResNet-50

ClickHouse

100M Rows Hits Dataset, First Run / Cold Cache

Neural Magic DeepSparse

Model: NLP Sentiment Analysis, 80% Pruned Quantized BERT Base Uncased - Scenario: Asynchronous Multi-Stream

Memcached

Set To Get Ratio: 1:100

Timed Godot Game Engine Compilation

Time To Compile

ONNX Runtime

Model: super-resolution-10 - Device: CPU - Executor: Standard

Darmstadt Automotive Parallel Heterogeneous Suite

Backend: OpenMP - Kernel: Euclidean Cluster

Timed LLVM Compilation

Build System: Unix Makefiles

TensorFlow

Device: CPU - Batch Size: 256 - Model: AlexNet

Zstd Compression

Compression Level: 8, Long Mode - Compression Speed

John The Ripper

Test: HMAC-SHA512

FFmpeg

Encoder: libx265 - Scenario: Live

FFmpeg

Encoder: libx265 - Scenario: Live

Neural Magic DeepSparse

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

Neural Magic DeepSparse

Model: CV Segmentation, 90% Pruned YOLACT Pruned - Scenario: Asynchronous Multi-Stream

ONNX Runtime

Model: Faster R-CNN R-50-FPN-int8 - Device: CPU - Executor: Standard

Neural Magic DeepSparse

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