Microsoft Azure EPYC 7003 HPC HBv3 Benchmarks
Azure HBv3 benchmarks against other Microsoft Azure HPC instance types. Benchmarks by Michael Larabel for a future article on phoronix.
HTML result view exported from: https://openbenchmarking.org/result/2104153-PTS-AZUREHPC42&sgm=1&ppd_QXp1cmUgSEJ2MyAtIEVQWUMgN1YxMw=2.70&ppd_QXp1cmUgSEJ2MiAtIEVQWUMgN1YxMg=2.70&ppd_QXp1cmUgSEJ2MSAtIEVQWUMgNzU1MQ=2.28&ppd_QXp1cmUgSEMgLSBYZW9uIDgxNjg=3.168&ppd_QXp1cmUgSCAtIFhlb24gRTUtMjY2NyB2Mw=1.599&ppt=DOLLAR&sgm=1&ppd_QXp1cmUgSEJ2MyAtIEVQWUMgN1YxMw=2.70&ppd_QXp1cmUgSEJ2MiAtIEVQWUMgN1YxMg=2.70&ppd_QXp1cmUgSEJ2MSAtIEVQWUMgNzU1MQ=2.28&ppd_QXp1cmUgSEMgLSBYZW9uIDgxNjg=3.168&ppd_QXp1cmUgSCAtIFhlb24gRTUtMjY2NyB2Mw=1.599&ppt=DOLLAR_PER_HOUR&sgm=1&hgv=Azure+HBv3+-+EPYC+7V13&ppd_QXp1cmUgSEJ2MyAtIEVQWUMgN1YxMw=2.70&ppd_QXp1cmUgSEJ2MiAtIEVQWUMgN1YxMg=2.70&ppd_QXp1cmUgSEJ2MSAtIEVQWUMgNzU1MQ=2.28&ppd_QXp1cmUgSEMgLSBYZW9uIDgxNjg=3.168&ppd_QXp1cmUgSCAtIFhlb24gRTUtMjY2NyB2Mw=1.599&ppt=DOLLAR_PER_HOUR&sgm=1&hgv=Azure+HBv3+-+EPYC+7V13&ppd_QXp1cmUgSEJ2MyAtIEVQWUMgN1YxMw=3.60&ppd_QXp1cmUgSEJ2MiAtIEVQWUMgN1YxMg=2.70&ppd_QXp1cmUgSEJ2MSAtIEVQWUMgNzU1MQ=2.28&ppd_QXp1cmUgSEMgLSBYZW9uIDgxNjg=3.168&ppd_QXp1cmUgSCAtIFhlb24gRTUtMjY2NyB2Mw=1.599&ppt=DOLLAR_PER_HOUR&sor&grs.
Rodinia
Test: OpenMP LavaMD
GROMACS
Water Benchmark
NAS Parallel Benchmarks
Test / Class: LU.C
LULESH
SVT-HEVC
Tuning: 1 - Input: Bosphorus 1080p
Zstd Compression
Compression Level: 8 - Compression Speed
Timed Node.js Compilation
Time To Compile
SVT-VP9
Tuning: Visual Quality Optimized - Input: Bosphorus 1080p
Xcompact3d Incompact3d
Input: X3D-benchmarking input.i3d
High Performance Conjugate Gradient
Timed LLVM Compilation
Time To Compile
SVT-AV1
Encoder Mode: Enc Mode 8 - Input: 1080p
Timed Linux Kernel Compilation
Time To Compile
PlaidML
FP16: No - Mode: Inference - Network: VGG19 - Device: CPU
PlaidML
FP16: No - Mode: Inference - Network: VGG16 - Device: CPU
Botan
Test: ChaCha20Poly1305
Botan
Test: ChaCha20Poly1305 - Decrypt
SVT-AV1
Encoder Mode: Enc Mode 0 - Input: 1080p
Pennant
Test: leblancbig
Pennant
Test: sedovbig
Zstd Compression
Compression Level: 19 - Decompression Speed
Zstd Compression
Compression Level: 19, Long Mode - Decompression Speed
Botan
Test: AES-256
Botan
Test: AES-256 - Decrypt
Zstd Compression
Compression Level: 8, Long Mode - Decompression Speed
FinanceBench
Benchmark: Bonds OpenMP
QuantLib
FinanceBench
Benchmark: Repo OpenMP
Rodinia
Test: OpenMP HotSpot3D
Zstd Compression
Compression Level: 8, Long Mode - Compression Speed
GNU GMP GMPbench
Total Time
Botan
Test: Twofish
Botan
Test: Twofish - Decrypt
Botan
Test: Blowfish - Decrypt
Botan
Test: Blowfish
TNN
Target: CPU - Model: SqueezeNet v1.1
Botan
Test: CAST-256
Botan
Test: KASUMI
Botan
Test: CAST-256 - Decrypt
Botan
Test: KASUMI - Decrypt
PlaidML
FP16: No - Mode: Inference - Network: ResNet 50 - Device: CPU
Mobile Neural Network
Model: inception-v3
Mobile Neural Network
Model: resnet-v2-50
Mobile Neural Network
Model: SqueezeNetV1.0
TensorFlow Lite
Model: SqueezeNet
oneDNN
Harness: Matrix Multiply Batch Shapes Transformer - Data Type: u8s8f32 - Engine: CPU
oneDNN
Harness: Recurrent Neural Network Inference - Data Type: bf16bf16bf16 - Engine: CPU
oneDNN
Harness: Recurrent Neural Network Training - Data Type: bf16bf16bf16 - Engine: CPU
oneDNN
Harness: Matrix Multiply Batch Shapes Transformer - Data Type: f32 - Engine: CPU
oneDNN
Harness: Recurrent Neural Network Inference - Data Type: u8s8f32 - Engine: CPU
oneDNN
Harness: Recurrent Neural Network Training - Data Type: u8s8f32 - Engine: CPU
oneDNN
Harness: Recurrent Neural Network Inference - Data Type: f32 - Engine: CPU
oneDNN
Harness: Recurrent Neural Network Training - Data Type: f32 - Engine: CPU
oneDNN
Harness: Deconvolution Batch shapes_3d - Data Type: u8s8f32 - Engine: CPU
oneDNN
Harness: Deconvolution Batch shapes_1d - Data Type: u8s8f32 - Engine: CPU
oneDNN
Harness: IP Shapes 3D - Data Type: u8s8f32 - Engine: CPU
oneDNN
Harness: IP Shapes 3D - Data Type: f32 - Engine: CPU
SVT-HEVC
Tuning: 10 - Input: Bosphorus 1080p
SVT-HEVC
Tuning: 7 - Input: Bosphorus 1080p
SVT-AV1
Encoder Mode: Enc Mode 4 - Input: 1080p
Zstd Compression
Compression Level: 19 - Compression Speed
NAMD
ATPase Simulation - 327,506 Atoms
CloverLeaf
Lagrangian-Eulerian Hydrodynamics
Geometric Mean Of All Test Results
Result Composite - Microsoft Azure EPYC 7003 HPC HBv3 Benchmarks
Rodinia
Performance / Cost - Test: OpenMP LavaMD
GROMACS
Performance / Cost - Water Benchmark
NAS Parallel Benchmarks
Performance / Cost - Test / Class: LU.C
LULESH
Performance / Cost
SVT-HEVC
Performance / Cost - Tuning: 1 - Input: Bosphorus 1080p
Zstd Compression
Performance / Cost - Compression Level: 8 - Compression Speed
Timed Node.js Compilation
Performance / Cost - Time To Compile
SVT-VP9
Performance / Cost - Tuning: Visual Quality Optimized - Input: Bosphorus 1080p
Xcompact3d Incompact3d
Performance / Cost - Input: X3D-benchmarking input.i3d
High Performance Conjugate Gradient
Performance / Cost
Timed LLVM Compilation
Performance / Cost - Time To Compile
SVT-AV1
Performance / Cost - Encoder Mode: Enc Mode 8 - Input: 1080p
Timed Linux Kernel Compilation
Performance / Cost - Time To Compile
PlaidML
Performance / Cost - FP16: No - Mode: Inference - Network: VGG19 - Device: CPU
PlaidML
Performance / Cost - FP16: No - Mode: Inference - Network: VGG16 - Device: CPU
Botan
Performance / Cost - Test: ChaCha20Poly1305
Botan
Performance / Cost - Test: ChaCha20Poly1305 - Decrypt
SVT-AV1
Performance / Cost - Encoder Mode: Enc Mode 0 - Input: 1080p
Pennant
Performance / Cost - Test: leblancbig
Pennant
Performance / Cost - Test: sedovbig
Zstd Compression
Performance / Cost - Compression Level: 19 - Decompression Speed
Zstd Compression
Performance / Cost - Compression Level: 19, Long Mode - Decompression Speed
Botan
Performance / Cost - Test: AES-256
Botan
Performance / Cost - Test: AES-256 - Decrypt
Zstd Compression
Performance / Cost - Compression Level: 8, Long Mode - Decompression Speed
FinanceBench
Performance / Cost - Benchmark: Bonds OpenMP
QuantLib
Performance / Cost
FinanceBench
Performance / Cost - Benchmark: Repo OpenMP
Rodinia
Performance / Cost - Test: OpenMP HotSpot3D
Zstd Compression
Performance / Cost - Compression Level: 8, Long Mode - Compression Speed
GNU GMP GMPbench
Performance / Cost - Total Time
Botan
Performance / Cost - Test: Twofish
Botan
Performance / Cost - Test: Twofish - Decrypt
Botan
Performance / Cost - Test: Blowfish - Decrypt
Botan
Performance / Cost - Test: Blowfish
TNN
Performance / Cost - Target: CPU - Model: SqueezeNet v1.1
Botan
Performance / Cost - Test: CAST-256
Botan
Performance / Cost - Test: KASUMI
Botan
Performance / Cost - Test: CAST-256 - Decrypt
Botan
Performance / Cost - Test: KASUMI - Decrypt
PlaidML
Performance / Cost - FP16: No - Mode: Inference - Network: ResNet 50 - Device: CPU
Mobile Neural Network
Performance / Cost - Model: inception-v3
Mobile Neural Network
Performance / Cost - Model: resnet-v2-50
Mobile Neural Network
Performance / Cost - Model: SqueezeNetV1.0
TensorFlow Lite
Performance / Cost - Model: SqueezeNet
oneDNN
Performance / Cost - Harness: Matrix Multiply Batch Shapes Transformer - Data Type: u8s8f32 - Engine: CPU
oneDNN
Performance / Cost - Harness: Recurrent Neural Network Inference - Data Type: bf16bf16bf16 - Engine: CPU
oneDNN
Performance / Cost - Harness: Recurrent Neural Network Training - Data Type: bf16bf16bf16 - Engine: CPU
oneDNN
Performance / Cost - Harness: Matrix Multiply Batch Shapes Transformer - Data Type: f32 - Engine: CPU
oneDNN
Performance / Cost - Harness: Recurrent Neural Network Inference - Data Type: u8s8f32 - Engine: CPU
oneDNN
Performance / Cost - Harness: Recurrent Neural Network Training - Data Type: u8s8f32 - Engine: CPU
oneDNN
Performance / Cost - Harness: Recurrent Neural Network Inference - Data Type: f32 - Engine: CPU
oneDNN
Performance / Cost - Harness: Recurrent Neural Network Training - Data Type: f32 - Engine: CPU
oneDNN
Performance / Cost - Harness: Deconvolution Batch shapes_3d - Data Type: u8s8f32 - Engine: CPU
oneDNN
Performance / Cost - Harness: Deconvolution Batch shapes_1d - Data Type: u8s8f32 - Engine: CPU
oneDNN
Performance / Cost - Harness: IP Shapes 3D - Data Type: u8s8f32 - Engine: CPU
oneDNN
Performance / Cost - Harness: IP Shapes 3D - Data Type: f32 - Engine: CPU
SVT-HEVC
Performance / Cost - Tuning: 10 - Input: Bosphorus 1080p
SVT-HEVC
Performance / Cost - Tuning: 7 - Input: Bosphorus 1080p
SVT-AV1
Performance / Cost - Encoder Mode: Enc Mode 4 - Input: 1080p
Zstd Compression
Performance / Cost - Compression Level: 19 - Compression Speed
NAMD
Performance / Cost - ATPase Simulation - 327,506 Atoms
CloverLeaf
Performance / Cost - Lagrangian-Eulerian Hydrodynamics
Phoronix Test Suite v10.8.5