Google Cloud c3 Sapphire Rapids vs. AMD Milan
Benchmarks by Michael Larabel for a future article.
HTML result view exported from: https://openbenchmarking.org/result/2303280-NE-2303286NE24&sgm=1&ppd_YzMtaGlnaGNwdS04IFNQUg=0.3448&ppd_YzItc3RhbmRhcmQtOCBDTFg=0.4176&ppd_bjItc3RhbmRhcmQtOCBDTFg=0.388472&ppd_bjItaGlnaGNwdS04IENMWA=0.286784&ppd_dDJkLXN0YW5kYXJkLTggQU1E=0.337968&ppd_YzJkLWhpZ2hjcHUtOCBBTUQ=0.2998&ppt=DPH&sgm=1&sts=1&ppd_YzMtaGlnaGNwdS04IFNQUg=0.3448&ppd_YzItc3RhbmRhcmQtOCBDTFg=0.4176&ppd_bjItc3RhbmRhcmQtOCBDTFg=0.388472&ppd_bjItaGlnaGNwdS04IENMWA=0.286784&ppd_dDJkLXN0YW5kYXJkLTggQU1E=0.337968&ppd_YzJkLWhpZ2hjcHUtOCBBTUQ=0.2998&ppt=DPH&sor.
LeelaChessZero
Backend: BLAS
LeelaChessZero
Backend: Eigen
miniBUDE
Implementation: OpenMP - Input Deck: BM1
miniBUDE
Implementation: OpenMP - Input Deck: BM1
NAMD
ATPase Simulation - 327,506 Atoms
nekRS
Input: TurboPipe Periodic
Xcompact3d Incompact3d
Input: input.i3d 129 Cells Per Direction
OpenFOAM
Input: drivaerFastback, Small Mesh Size - Mesh Time
OpenFOAM
Input: drivaerFastback, Small Mesh Size - Execution Time
OpenRadioss
Model: Bumper Beam
OpenRadioss
Model: Cell Phone Drop Test
OpenRadioss
Model: Bird Strike on Windshield
OpenRadioss
Model: Rubber O-Ring Seal Installation
SPECFEM3D
Model: Mount St. Helens
SPECFEM3D
Model: Layered Halfspace
SPECFEM3D
Model: Tomographic Model
SPECFEM3D
Model: Homogeneous Halfspace
SPECFEM3D
Model: Water-layered Halfspace
Zstd Compression
Compression Level: 19 - Compression Speed
Zstd Compression
Compression Level: 19 - Decompression Speed
Zstd Compression
Compression Level: 19, Long Mode - Compression Speed
Zstd Compression
Compression Level: 19, Long Mode - Decompression Speed
John The Ripper
Test: bcrypt
John The Ripper
Test: Blowfish
John The Ripper
Test: MD5
Embree
Binary: Pathtracer ISPC - Model: Crown
Embree
Binary: Pathtracer ISPC - Model: Asian Dragon
uvg266
Video Input: Bosphorus 4K - Video Preset: Very Fast
uvg266
Video Input: Bosphorus 4K - Video Preset: Super Fast
uvg266
Video Input: Bosphorus 4K - Video Preset: Ultra Fast
uvg266
Video Input: Bosphorus 1080p - Video Preset: Very Fast
uvg266
Video Input: Bosphorus 1080p - Video Preset: Super Fast
uvg266
Video Input: Bosphorus 1080p - Video Preset: Ultra Fast
Intel Open Image Denoise
Run: RT.hdr_alb_nrm.3840x2160
Intel Open Image Denoise
Run: RTLightmap.hdr.4096x4096
OpenVKL
Benchmark: vklBenchmark ISPC
7-Zip Compression
Test: Compression Rating
Timed FFmpeg Compilation
Time To Compile
Timed Linux Kernel Compilation
Build: defconfig
oneDNN
Harness: IP Shapes 1D - Data Type: bf16bf16bf16 - Engine: CPU
oneDNN
Harness: IP Shapes 3D - Data Type: bf16bf16bf16 - Engine: CPU
oneDNN
Harness: Convolution Batch Shapes Auto - Data Type: bf16bf16bf16 - Engine: CPU
oneDNN
Harness: Deconvolution Batch shapes_1d - Data Type: bf16bf16bf16 - Engine: CPU
oneDNN
Harness: Deconvolution Batch shapes_3d - Data Type: bf16bf16bf16 - Engine: CPU
oneDNN
Harness: Recurrent Neural Network Training - Data Type: bf16bf16bf16 - Engine: CPU
oneDNN
Harness: Recurrent Neural Network Inference - Data Type: bf16bf16bf16 - Engine: CPU
oneDNN
Harness: Matrix Multiply Batch Shapes Transformer - Data Type: bf16bf16bf16 - Engine: CPU
OSPRay Studio
Camera: 1 - Resolution: 4K - Samples Per Pixel: 1 - Renderer: Path Tracer
OSPRay Studio
Camera: 3 - Resolution: 4K - Samples Per Pixel: 1 - Renderer: Path Tracer
OpenSSL
Algorithm: SHA256
OpenSSL
Algorithm: SHA512
OpenSSL
Algorithm: RSA4096
OpenSSL
Algorithm: ChaCha20
OpenSSL
Algorithm: AES-128-GCM
OpenSSL
Algorithm: AES-256-GCM
OpenSSL
Algorithm: ChaCha20-Poly1305
CockroachDB
Workload: KV, 50% Reads - Concurrency: 128
CockroachDB
Workload: KV, 95% Reads - Concurrency: 128
Memcached
Set To Get Ratio: 1:10
Memcached
Set To Get Ratio: 1:100
GROMACS
Implementation: MPI CPU - Input: water_GMX50_bare
MariaDB
Clients: 2048
MariaDB
Clients: 4096
PostgreSQL
Scaling Factor: 100 - Clients: 800 - Mode: Read Only
PostgreSQL
Scaling Factor: 100 - Clients: 800 - Mode: Read Only - Average Latency
PostgreSQL
Scaling Factor: 100 - Clients: 1000 - Mode: Read Only
PostgreSQL
Scaling Factor: 100 - Clients: 1000 - Mode: Read Only - Average Latency
TensorFlow
Device: CPU - Batch Size: 16 - Model: ResNet-50
TensorFlow
Device: CPU - Batch Size: 32 - Model: ResNet-50
TensorFlow
Device: CPU - Batch Size: 64 - Model: ResNet-50
Neural Magic DeepSparse
Model: NLP Document Classification, oBERT base uncased on IMDB - Scenario: Asynchronous Multi-Stream
Neural Magic DeepSparse
Model: NLP Document Classification, oBERT base uncased on IMDB - Scenario: Asynchronous Multi-Stream
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
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
Neural Magic DeepSparse
Model: NLP Text Classification, DistilBERT mnli - Scenario: Asynchronous Multi-Stream
Neural Magic DeepSparse
Model: NLP Text Classification, DistilBERT mnli - Scenario: Asynchronous Multi-Stream
Neural Magic DeepSparse
Model: CV Segmentation, 90% Pruned YOLACT Pruned - Scenario: Asynchronous Multi-Stream
Neural Magic DeepSparse
Model: CV Segmentation, 90% Pruned YOLACT Pruned - 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, BERT base uncased SST2 - Scenario: Asynchronous Multi-Stream
Neural Magic DeepSparse
Model: NLP Token Classification, BERT base uncased conll2003 - Scenario: Asynchronous Multi-Stream
Neural Magic DeepSparse
Model: NLP Token Classification, BERT base uncased conll2003 - Scenario: Asynchronous Multi-Stream
Google Draco
Model: Lion
Google Draco
Model: Church Facade
Blender
Blend File: BMW27 - Compute: CPU-Only
nginx
Connections: 100
nginx
Connections: 200
nginx
Connections: 500
nginx
Connections: 1000
nginx
Connections: 4000
BRL-CAD
VGR Performance Metric
OpenCV
Test: Core
OpenCV
Test: Graph API
OpenCV
Test: Stitching
OpenCV
Test: Image Processing
OpenCV
Test: Object Detection
Geometric Mean Of All Test Results
Result Composite - Google Cloud c3 Sapphire Rapids vs. AMD Milan
Geometric Mean Of Timed Code Compilation Tests
Result Composite - Google Cloud c3 Sapphire Rapids vs. AMD Milan
Geometric Mean Of C/C++ Compiler Tests
Result Composite - Google Cloud c3 Sapphire Rapids vs. AMD Milan
Geometric Mean Of Compression Tests
Result Composite - Google Cloud c3 Sapphire Rapids vs. AMD Milan
Geometric Mean Of CPU Massive Tests
Result Composite - Google Cloud c3 Sapphire Rapids vs. AMD Milan
Geometric Mean Of Creator Workloads Tests
Result Composite - Google Cloud c3 Sapphire Rapids vs. AMD Milan
Geometric Mean Of Cryptography Tests
Result Composite - Google Cloud c3 Sapphire Rapids vs. AMD Milan
Geometric Mean Of Database Test Suite
Result Composite - Google Cloud c3 Sapphire Rapids vs. AMD Milan
Geometric Mean Of Fortran Tests
Result Composite - Google Cloud c3 Sapphire Rapids vs. AMD Milan
Geometric Mean Of Game Development Tests
Result Composite - Google Cloud c3 Sapphire Rapids vs. AMD Milan
Geometric Mean Of HPC - High Performance Computing Tests
Result Composite - Google Cloud c3 Sapphire Rapids vs. AMD Milan
Geometric Mean Of Common Kernel Benchmarks Tests
Result Composite - Google Cloud c3 Sapphire Rapids vs. AMD Milan
Geometric Mean Of Machine Learning Tests
Result Composite - Google Cloud c3 Sapphire Rapids vs. AMD Milan
Geometric Mean Of Molecular Dynamics Tests
Result Composite - Google Cloud c3 Sapphire Rapids vs. AMD Milan
Geometric Mean Of MPI Benchmarks Tests
Result Composite - Google Cloud c3 Sapphire Rapids vs. AMD Milan
Geometric Mean Of Multi-Core Tests
Result Composite - Google Cloud c3 Sapphire Rapids vs. AMD Milan
Geometric Mean Of NVIDIA GPU Compute Tests
Result Composite - Google Cloud c3 Sapphire Rapids vs. AMD Milan
Geometric Mean Of Intel oneAPI Tests
Result Composite - Google Cloud c3 Sapphire Rapids vs. AMD Milan
Geometric Mean Of OpenMPI Tests
Result Composite - Google Cloud c3 Sapphire Rapids vs. AMD Milan
Geometric Mean Of Programmer / Developer System Benchmarks Tests
Result Composite - Google Cloud c3 Sapphire Rapids vs. AMD Milan
Geometric Mean Of Python Tests
Result Composite - Google Cloud c3 Sapphire Rapids vs. AMD Milan
Geometric Mean Of Renderers Tests
Result Composite - Google Cloud c3 Sapphire Rapids vs. AMD Milan
Geometric Mean Of Scientific Computing Tests
Result Composite - Google Cloud c3 Sapphire Rapids vs. AMD Milan
Geometric Mean Of Server Tests
Result Composite - Google Cloud c3 Sapphire Rapids vs. AMD Milan
Geometric Mean Of Server CPU Tests
Result Composite - Google Cloud c3 Sapphire Rapids vs. AMD Milan
LeelaChessZero
Performance / Cost - Backend: BLAS
LeelaChessZero
Performance / Cost - Backend: Eigen
miniBUDE
Performance / Cost - Implementation: OpenMP - Input Deck: BM1
miniBUDE
Performance / Cost - Implementation: OpenMP - Input Deck: BM1
NAMD
Performance / Cost - ATPase Simulation - 327,506 Atoms
nekRS
Performance / Cost - Input: TurboPipe Periodic
Xcompact3d Incompact3d
Performance / Cost - Input: input.i3d 129 Cells Per Direction
OpenFOAM
Performance / Cost - Input: drivaerFastback, Small Mesh Size - Mesh Time
OpenFOAM
Performance / Cost - Input: drivaerFastback, Small Mesh Size - Execution Time
OpenRadioss
Performance / Cost - Model: Bumper Beam
OpenRadioss
Performance / Cost - Model: Cell Phone Drop Test
OpenRadioss
Performance / Cost - Model: Bird Strike on Windshield
OpenRadioss
Performance / Cost - Model: Rubber O-Ring Seal Installation
SPECFEM3D
Performance / Cost - Model: Mount St. Helens
SPECFEM3D
Performance / Cost - Model: Layered Halfspace
SPECFEM3D
Performance / Cost - Model: Tomographic Model
SPECFEM3D
Performance / Cost - Model: Homogeneous Halfspace
SPECFEM3D
Performance / Cost - Model: Water-layered Halfspace
Zstd Compression
Performance / Cost - Compression Level: 19 - Compression Speed
Zstd Compression
Performance / Cost - Compression Level: 19 - Decompression Speed
Zstd Compression
Performance / Cost - Compression Level: 19, Long Mode - Compression Speed
Zstd Compression
Performance / Cost - Compression Level: 19, Long Mode - Decompression Speed
John The Ripper
Performance / Cost - Test: bcrypt
John The Ripper
Performance / Cost - Test: Blowfish
John The Ripper
Performance / Cost - Test: MD5
Embree
Performance / Cost - Binary: Pathtracer ISPC - Model: Crown
Embree
Performance / Cost - Binary: Pathtracer ISPC - Model: Asian Dragon
uvg266
Performance / Cost - Video Input: Bosphorus 4K - Video Preset: Very Fast
uvg266
Performance / Cost - Video Input: Bosphorus 4K - Video Preset: Super Fast
uvg266
Performance / Cost - Video Input: Bosphorus 4K - Video Preset: Ultra Fast
uvg266
Performance / Cost - Video Input: Bosphorus 1080p - Video Preset: Very Fast
uvg266
Performance / Cost - Video Input: Bosphorus 1080p - Video Preset: Super Fast
uvg266
Performance / Cost - Video Input: Bosphorus 1080p - Video Preset: Ultra Fast
Intel Open Image Denoise
Performance / Cost - Run: RT.hdr_alb_nrm.3840x2160
Intel Open Image Denoise
Performance / Cost - Run: RTLightmap.hdr.4096x4096
OpenVKL
Performance / Cost - Benchmark: vklBenchmark ISPC
7-Zip Compression
Performance / Cost - Test: Compression Rating
Timed FFmpeg Compilation
Performance / Cost - Time To Compile
Timed Linux Kernel Compilation
Performance / Cost - Build: defconfig
oneDNN
Performance / Cost - Harness: IP Shapes 1D - Data Type: bf16bf16bf16 - Engine: CPU
oneDNN
Performance / Cost - Harness: IP Shapes 3D - Data Type: bf16bf16bf16 - Engine: CPU
oneDNN
Performance / Cost - Harness: Convolution Batch Shapes Auto - Data Type: bf16bf16bf16 - Engine: CPU
oneDNN
Performance / Cost - Harness: Deconvolution Batch shapes_1d - Data Type: bf16bf16bf16 - Engine: CPU
oneDNN
Performance / Cost - Harness: Recurrent Neural Network Training - Data Type: bf16bf16bf16 - Engine: CPU
oneDNN
Performance / Cost - Harness: Recurrent Neural Network Inference - Data Type: bf16bf16bf16 - Engine: CPU
oneDNN
Performance / Cost - Harness: Matrix Multiply Batch Shapes Transformer - Data Type: bf16bf16bf16 - Engine: CPU
OSPRay Studio
Performance / Cost - Camera: 1 - Resolution: 4K - Samples Per Pixel: 1 - Renderer: Path Tracer
OSPRay Studio
Performance / Cost - Camera: 3 - Resolution: 4K - Samples Per Pixel: 1 - Renderer: Path Tracer
OpenSSL
Performance / Cost - Algorithm: SHA256
OpenSSL
Performance / Cost - Algorithm: SHA512
OpenSSL
Performance / Cost - Algorithm: RSA4096
OpenSSL
Performance / Cost - Algorithm: ChaCha20
OpenSSL
Performance / Cost - Algorithm: AES-128-GCM
OpenSSL
Performance / Cost - Algorithm: AES-256-GCM
OpenSSL
Performance / Cost - Algorithm: ChaCha20-Poly1305
CockroachDB
Performance / Cost - Workload: KV, 50% Reads - Concurrency: 128
CockroachDB
Performance / Cost - Workload: KV, 95% Reads - Concurrency: 128
Memcached
Performance / Cost - Set To Get Ratio: 1:10
Memcached
Performance / Cost - Set To Get Ratio: 1:100
GROMACS
Performance / Cost - Implementation: MPI CPU - Input: water_GMX50_bare
MariaDB
Performance / Cost - Clients: 2048
MariaDB
Performance / Cost - Clients: 4096
PostgreSQL
Performance / Cost - Scaling Factor: 100 - Clients: 800 - Mode: Read Only
PostgreSQL
Performance / Cost - Scaling Factor: 100 - Clients: 800 - Mode: Read Only - Average Latency
PostgreSQL
Performance / Cost - Scaling Factor: 100 - Clients: 1000 - Mode: Read Only
PostgreSQL
Performance / Cost - Scaling Factor: 100 - Clients: 1000 - Mode: Read Only - Average Latency
TensorFlow
Performance / Cost - Device: CPU - Batch Size: 16 - Model: ResNet-50
TensorFlow
Performance / Cost - Device: CPU - Batch Size: 32 - Model: ResNet-50
TensorFlow
Performance / Cost - Device: CPU - Batch Size: 64 - Model: ResNet-50
Neural Magic DeepSparse
Performance / Cost - Model: NLP Document Classification, oBERT base uncased on IMDB - Scenario: Asynchronous Multi-Stream
Neural Magic DeepSparse
Performance / Cost - Model: NLP Document Classification, oBERT base uncased on IMDB - Scenario: Asynchronous Multi-Stream
Neural Magic DeepSparse
Performance / Cost - Model: NLP Question Answering, BERT base uncased SQuaD 12layer Pruned90 - Scenario: Asynchronous Multi-Stream
Neural Magic DeepSparse
Performance / Cost - Model: NLP Question Answering, BERT base uncased SQuaD 12layer Pruned90 - Scenario: Asynchronous Multi-Stream
Neural Magic DeepSparse
Performance / Cost - Model: CV Classification, ResNet-50 ImageNet - Scenario: Asynchronous Multi-Stream
Neural Magic DeepSparse
Performance / Cost - Model: CV Classification, ResNet-50 ImageNet - Scenario: Asynchronous Multi-Stream
Neural Magic DeepSparse
Performance / Cost - Model: NLP Text Classification, DistilBERT mnli - Scenario: Asynchronous Multi-Stream
Neural Magic DeepSparse
Performance / Cost - Model: NLP Text Classification, DistilBERT mnli - Scenario: Asynchronous Multi-Stream
Neural Magic DeepSparse
Performance / Cost - Model: CV Segmentation, 90% Pruned YOLACT Pruned - Scenario: Asynchronous Multi-Stream
Neural Magic DeepSparse
Performance / Cost - Model: CV Segmentation, 90% Pruned YOLACT Pruned - Scenario: Asynchronous Multi-Stream
Neural Magic DeepSparse
Performance / Cost - Model: NLP Text Classification, BERT base uncased SST2 - Scenario: Asynchronous Multi-Stream
Neural Magic DeepSparse
Performance / Cost - Model: NLP Text Classification, BERT base uncased SST2 - Scenario: Asynchronous Multi-Stream
Neural Magic DeepSparse
Performance / Cost - Model: NLP Token Classification, BERT base uncased conll2003 - Scenario: Asynchronous Multi-Stream
Neural Magic DeepSparse
Performance / Cost - Model: NLP Token Classification, BERT base uncased conll2003 - Scenario: Asynchronous Multi-Stream
Google Draco
Performance / Cost - Model: Lion
Google Draco
Performance / Cost - Model: Church Facade
Blender
Performance / Cost - Blend File: BMW27 - Compute: CPU-Only
nginx
Performance / Cost - Connections: 100
nginx
Performance / Cost - Connections: 200
nginx
Performance / Cost - Connections: 500
nginx
Performance / Cost - Connections: 1000
nginx
Performance / Cost - Connections: 4000
BRL-CAD
Performance / Cost - VGR Performance Metric
OpenCV
Performance / Cost - Test: Core
OpenCV
Performance / Cost - Test: Graph API
OpenCV
Performance / Cost - Test: Stitching
OpenCV
Performance / Cost - Test: Image Processing
OpenCV
Performance / Cost - Test: Object Detection
Phoronix Test Suite v10.8.5