Amazon EC2 Graviton3 Benchmark Comparison
Amazon AWS Graviton3 benchmarks by Michael Larabel.
HTML result view exported from: https://openbenchmarking.org/result/2205260-PTS-GRAVITON42&sgm=1&swl=1&ppd_YTEuNHhsYXJnZSBHcmF2aXRvbg=0.408&ppd_YzZnLjR4bGFyZ2UgR3Jhdml0b24y=0.544&ppd_YzdnLjR4bGFyZ2UgR3Jhdml0b24z=0.58&ppd_YzZhLjR4bGFyZ2UgRVBZQw=0.612&ppd_YzZpLjR4bGFyZ2UgWGVvbg=0.68&ppt=DPH&grs&rdt.
Stress-NG
Test: Memory Copying
NAS Parallel Benchmarks
Test / Class: MG.C
NAS Parallel Benchmarks
Test / Class: CG.C
NAS Parallel Benchmarks
Test / Class: SP.C
Zstd Compression
Compression Level: 3 - Compression Speed
NAS Parallel Benchmarks
Test / Class: FT.C
High Performance Conjugate Gradient
Algebraic Multi-Grid Benchmark
Xcompact3d Incompact3d
Input: input.i3d 129 Cells Per Direction
ACES DGEMM
Sustained Floating-Point Rate
Xcompact3d Incompact3d
Input: input.i3d 193 Cells Per Direction
Stress-NG
Test: CPU Stress
simdjson
Throughput Test: DistinctUserID
Timed MrBayes Analysis
Primate Phylogeny Analysis
NAS Parallel Benchmarks
Test / Class: IS.D
TensorFlow Lite
Model: Mobilenet Float
GPAW
Input: Carbon Nanotube
libavif avifenc
Encoder Speed: 2
simdjson
Throughput Test: PartialTweets
LULESH
Apache HTTP Server
Concurrent Requests: 100
ASTC Encoder
Preset: Thorough
GROMACS
Implementation: MPI CPU - Input: water_GMX50_bare
TensorFlow Lite
Model: Inception V4
Apache HTTP Server
Concurrent Requests: 500
Apache HTTP Server
Concurrent Requests: 200
simdjson
Throughput Test: Kostya
NAS Parallel Benchmarks
Test / Class: BT.C
OpenSSL
Algorithm: RSA4096
TensorFlow Lite
Model: Inception ResNet V2
Apache HTTP Server
Concurrent Requests: 1000
TensorFlow Lite
Model: SqueezeNet
ASTC Encoder
Preset: Exhaustive
Rodinia
Test: OpenMP CFD Solver
OpenSSL
Algorithm: RSA4096
libavif avifenc
Encoder Speed: 0
Timed Node.js Compilation
Time To Compile
LAMMPS Molecular Dynamics Simulator
Model: Rhodopsin Protein
PyBench
Total For Average Test Times
PHPBench
PHP Benchmark Suite
Timed ImageMagick Compilation
Time To Compile
TensorFlow Lite
Model: NASNet Mobile
Timed Apache Compilation
Time To Compile
DaCapo Benchmark
Java Test: Jython
Timed LLVM Compilation
Build System: Ninja
NAS Parallel Benchmarks
Test / Class: EP.D
Ngspice
Circuit: C2670
DaCapo Benchmark
Java Test: Tradesoap
simdjson
Throughput Test: LargeRandom
SecureMark
Benchmark: SecureMark-TLS
DaCapo Benchmark
Java Test: Tradebeans
Liquid-DSP
Threads: 16 - Buffer Length: 256 - Filter Length: 57
Build2
Time To Compile
Timed PHP Compilation
Time To Compile
7-Zip Compression
Test: Compression Rating
Ngspice
Circuit: C7552
WebP Image Encode
Encode Settings: Quality 100, Lossless, Highest Compression
Timed Gem5 Compilation
Time To Compile
WebP Image Encode
Encode Settings: Quality 100, Lossless
libavif avifenc
Encoder Speed: 6, Lossless
nginx
Concurrent Requests: 1000
nginx
Concurrent Requests: 500
nginx
Concurrent Requests: 200
Zstd Compression
Compression Level: 19 - Decompression Speed
nginx
Concurrent Requests: 100
TSCP
AI Chess Performance
Zstd Compression
Compression Level: 19, Long Mode - Decompression Speed
Stockfish
Total Time
Rodinia
Test: OpenMP LavaMD
POV-Ray
Trace Time
Zstd Compression
Compression Level: 19, Long Mode - Compression Speed
Zstd Compression
Compression Level: 19 - Compression Speed
DaCapo Benchmark
Java Test: H2
Stress-NG
Test: Crypto
asmFish
1024 Hash Memory, 26 Depth
Google SynthMark
Test: VoiceMark_100
OpenSSL
Algorithm: SHA256
Stress-NG
Test: Vector Math
NAS Parallel Benchmarks
Test / Class: LU.C
LeelaChessZero
Backend: Eigen
LeelaChessZero
Backend: BLAS
Coremark
CoreMark Size 666 - Iterations Per Second
N-Queens
Elapsed Time
7-Zip Compression
Test: Decompression Rating
m-queens
Time To Solve
Stress-NG
Test: IO_uring
ONNX Runtime
Model: super-resolution-10 - Device: CPU - Executor: Standard
ONNX Runtime
Model: ArcFace ResNet-100 - Device: CPU - Executor: Standard
ONNX Runtime
Model: fcn-resnet101-11 - Device: CPU - Executor: Standard
ONNX Runtime
Model: bertsquad-12 - Device: CPU - Executor: Standard
ONNX Runtime
Model: GPT-2 - Device: CPU - Executor: Standard
TensorFlow Lite
Model: Mobilenet Quant
C-Ray
Total Time - 4K, 16 Rays Per Pixel
Rodinia
Test: OpenMP Streamcluster
Geometric Mean Of All Test Results
Result Composite - Amazon EC2 Graviton3 Benchmark Comparison
Number Of First Place Finishes
Wins - 94 Tests
Number Of Last Place Finishes
Losses - 94 Tests
Stress-NG
Performance / Cost - Test: Memory Copying
NAS Parallel Benchmarks
Performance / Cost - Test / Class: MG.C
NAS Parallel Benchmarks
Performance / Cost - Test / Class: CG.C
NAS Parallel Benchmarks
Performance / Cost - Test / Class: SP.C
Zstd Compression
Performance / Cost - Compression Level: 3 - Compression Speed
NAS Parallel Benchmarks
Performance / Cost - Test / Class: FT.C
High Performance Conjugate Gradient
Performance / Cost
Algebraic Multi-Grid Benchmark
Performance / Cost
Xcompact3d Incompact3d
Performance / Cost - Input: input.i3d 129 Cells Per Direction
ACES DGEMM
Performance / Cost - Sustained Floating-Point Rate
Xcompact3d Incompact3d
Performance / Cost - Input: input.i3d 193 Cells Per Direction
Stress-NG
Performance / Cost - Test: CPU Stress
simdjson
Performance / Cost - Throughput Test: DistinctUserID
Timed MrBayes Analysis
Performance / Cost - Primate Phylogeny Analysis
NAS Parallel Benchmarks
Performance / Cost - Test / Class: IS.D
TensorFlow Lite
Performance / Cost - Model: Mobilenet Float
GPAW
Performance / Cost - Input: Carbon Nanotube
libavif avifenc
Performance / Cost - Encoder Speed: 2
simdjson
Performance / Cost - Throughput Test: PartialTweets
LULESH
Performance / Cost
Apache HTTP Server
Performance / Cost - Concurrent Requests: 100
ASTC Encoder
Performance / Cost - Preset: Thorough
GROMACS
Performance / Cost - Implementation: MPI CPU - Input: water_GMX50_bare
TensorFlow Lite
Performance / Cost - Model: Inception V4
Apache HTTP Server
Performance / Cost - Concurrent Requests: 500
Apache HTTP Server
Performance / Cost - Concurrent Requests: 200
simdjson
Performance / Cost - Throughput Test: Kostya
NAS Parallel Benchmarks
Performance / Cost - Test / Class: BT.C
OpenSSL
Performance / Cost - Algorithm: RSA4096
TensorFlow Lite
Performance / Cost - Model: Inception ResNet V2
Apache HTTP Server
Performance / Cost - Concurrent Requests: 1000
TensorFlow Lite
Performance / Cost - Model: SqueezeNet
ASTC Encoder
Performance / Cost - Preset: Exhaustive
Rodinia
Performance / Cost - Test: OpenMP CFD Solver
OpenSSL
Performance / Cost - Algorithm: RSA4096
libavif avifenc
Performance / Cost - Encoder Speed: 0
Timed Node.js Compilation
Performance / Cost - Time To Compile
LAMMPS Molecular Dynamics Simulator
Performance / Cost - Model: Rhodopsin Protein
PyBench
Performance / Cost - Total For Average Test Times
PHPBench
Performance / Cost - PHP Benchmark Suite
Timed ImageMagick Compilation
Performance / Cost - Time To Compile
TensorFlow Lite
Performance / Cost - Model: NASNet Mobile
Timed Apache Compilation
Performance / Cost - Time To Compile
DaCapo Benchmark
Performance / Cost - Java Test: Jython
Timed LLVM Compilation
Performance / Cost - Build System: Ninja
NAS Parallel Benchmarks
Performance / Cost - Test / Class: EP.D
Ngspice
Performance / Cost - Circuit: C2670
DaCapo Benchmark
Performance / Cost - Java Test: Tradesoap
simdjson
Performance / Cost - Throughput Test: LargeRandom
SecureMark
Performance / Cost - Benchmark: SecureMark-TLS
DaCapo Benchmark
Performance / Cost - Java Test: Tradebeans
Liquid-DSP
Performance / Cost - Threads: 16 - Buffer Length: 256 - Filter Length: 57
Build2
Performance / Cost - Time To Compile
Timed PHP Compilation
Performance / Cost - Time To Compile
7-Zip Compression
Performance / Cost - Test: Compression Rating
Ngspice
Performance / Cost - Circuit: C7552
WebP Image Encode
Performance / Cost - Encode Settings: Quality 100, Lossless, Highest Compression
Timed Gem5 Compilation
Performance / Cost - Time To Compile
WebP Image Encode
Performance / Cost - Encode Settings: Quality 100, Lossless
libavif avifenc
Performance / Cost - Encoder Speed: 6, Lossless
nginx
Performance / Cost - Concurrent Requests: 1000
nginx
Performance / Cost - Concurrent Requests: 500
nginx
Performance / Cost - Concurrent Requests: 200
Zstd Compression
Performance / Cost - Compression Level: 19 - Decompression Speed
nginx
Performance / Cost - Concurrent Requests: 100
Zstd Compression
Performance / Cost - Compression Level: 19, Long Mode - Decompression Speed
Stockfish
Performance / Cost - Total Time
Rodinia
Performance / Cost - Test: OpenMP LavaMD
POV-Ray
Performance / Cost - Trace Time
Zstd Compression
Performance / Cost - Compression Level: 19, Long Mode - Compression Speed
Zstd Compression
Performance / Cost - Compression Level: 19 - Compression Speed
DaCapo Benchmark
Performance / Cost - Java Test: H2
Stress-NG
Performance / Cost - Test: Crypto
asmFish
Performance / Cost - 1024 Hash Memory, 26 Depth
Google SynthMark
Performance / Cost - Test: VoiceMark_100
OpenSSL
Performance / Cost - Algorithm: SHA256
Stress-NG
Performance / Cost - Test: Vector Math
NAS Parallel Benchmarks
Performance / Cost - Test / Class: LU.C
LeelaChessZero
Performance / Cost - Backend: Eigen
LeelaChessZero
Performance / Cost - Backend: BLAS
Coremark
Performance / Cost - CoreMark Size 666 - Iterations Per Second
N-Queens
Performance / Cost - Elapsed Time
7-Zip Compression
Performance / Cost - Test: Decompression Rating
m-queens
Performance / Cost - Time To Solve
Stress-NG
Performance / Cost - Test: IO_uring
ONNX Runtime
Performance / Cost - Model: super-resolution-10 - Device: CPU - Executor: Standard
ONNX Runtime
Performance / Cost - Model: ArcFace ResNet-100 - Device: CPU - Executor: Standard
ONNX Runtime
Performance / Cost - Model: fcn-resnet101-11 - Device: CPU - Executor: Standard
ONNX Runtime
Performance / Cost - Model: bertsquad-12 - Device: CPU - Executor: Standard
ONNX Runtime
Performance / Cost - Model: GPT-2 - Device: CPU - Executor: Standard
TensorFlow Lite
Performance / Cost - Model: Mobilenet Quant
C-Ray
Performance / Cost - Total Time - 4K, 16 Rays Per Pixel
Rodinia
Performance / Cost - Test: OpenMP Streamcluster
Phoronix Test Suite v10.8.5