NVIDIA Jetson Nano Benchmarks

ARMv8 rev 1 testing with a NVIDIA Jetson Nano Developer Kit and NVIDIA TEGRA on Ubuntu 18.04 via the Phoronix Test Suite.

HTML result view exported from: https://openbenchmarking.org/result/1908084-HV-1903316HV84&grw.

NVIDIA Jetson Nano BenchmarksProcessorMotherboardMemoryDiskGraphicsMonitorNetworkOSKernelDesktopDisplay ServerDisplay DriverOpenGLVulkanCompilerFile-SystemScreen ResolutionJetson NanoNanonanoNano 5WNano 10wARMv8 rev 1 @ 1.43GHz (4 Cores)jetson-nano4096MB32GB GB1QTNVIDIA Tegra X1VE228Realtek RTL8111/8168/8411Ubuntu 18.044.9.140-tegra (aarch64)Unity 7.5.0X Server 1.19.6NVIDIA 32.1.04.6.01.1.85GCC 7.3.0 + CUDA 10.0ext41920x1080ARMv8 rev 1 @ 0.92GHz (2 Cores)NVIDIA Jetson Nano Developer Kit64GB SN64GNVIDIA TEGRANVIDIA 1.0.0GCC 7.4.0 + CUDA 10.0ARMv8 rev 1 @ 1.43GHz (4 Cores)OpenBenchmarking.orgCompiler Details- --build=aarch64-linux-gnu --disable-libquadmath --disable-libquadmath-support --disable-werror --enable-checking=release --enable-clocale=gnu --enable-default-pie --enable-fix-cortex-a53-843419 --enable-gnu-unique-object --enable-languages=c,ada,c++,go,d,fortran,objc,obj-c++ --enable-libstdcxx-debug --enable-libstdcxx-time=yes --enable-multiarch --enable-nls --enable-plugin --enable-shared --enable-threads=posix --host=aarch64-linux-gnu --program-prefix=aarch64-linux-gnu- --target=aarch64-linux-gnu --with-default-libstdcxx-abi=new --with-gcc-major-version-only -v Processor Details- Scaling Governor: tegra-cpufreq schedutilPython Details- Jetson Nano: Python 2.7.15rc1 + Python 3.6.7- Nano: Python 2.7.15+ + Python 3.6.8- nano: Python 2.7.15+ + Python 3.6.8- Nano 5W: Python 2.7.15+ + Python 3.6.8- Nano 10w: Python 2.7.15+ + Python 3.6.8Java Details- Nano: OpenJDK Runtime Environment (build 11.0.3+7-Ubuntu-1ubuntu218.04.1)- nano: OpenJDK Runtime Environment (build 11.0.3+7-Ubuntu-1ubuntu218.04.1)- Nano 5W: OpenJDK Runtime Environment (build 11.0.3+7-Ubuntu-1ubuntu218.04.1)- Nano 10w: OpenJDK Runtime Environment (build 11.0.4+11-post-Ubuntu-1ubuntu218.04.3)

NVIDIA Jetson Nano Benchmarkscuda-mini-nbody: SOA Data Layoutcuda-mini-nbody: Flush Denormals To Zerocuda-mini-nbody: Loop Unrollingtensorrt-inference: GoogleNet - INT8 - 1 - Disabledcompress-xz: Compressing ubuntu-16.04.3-server-i386.img, Compression Level 9mbw: Memory Copy - 128 MiBtensorrt-inference: ResNet50 - INT8 - 8 - Disabledmbw: Memory Copy - 512 MiBtensorrt-inference: AlexNet - INT8 - 32 - Disabledtensorrt-inference: GoogleNet - INT8 - 4 - Disabledtensorrt-inference: AlexNet - INT8 - 16 - Disabledtensorrt-inference: VGG16 - FP16 - 1 - Disabledtensorrt-inference: AlexNet - FP16 - 16 - Disabledtensorrt-inference: ResNet50 - FP16 - 8 - Disabledtensorrt-inference: ResNet50 - INT8 - 4 - Disabledtensorrt-inference: AlexNet - FP16 - 4 - Disabledtensorrt-inference: AlexNet - FP16 - 8 - Disabledtensorrt-inference: VGG19 - FP16 - 4 - Disabledtensorrt-inference: AlexNet - FP16 - 1 - Disabledtensorrt-inference: VGG19 - FP16 - 1 - Disabledtensorrt-inference: VGG16 - FP16 - 8 - Disabledtensorrt-inference: VGG16 - FP16 - 4 - Disabledtensorrt-inference: ResNet50 - FP16 - 1 - Disabledtensorrt-inference: AlexNet - INT8 - 1 - Disabledtensorrt-inference: ResNet50 - FP16 - 4 - Disabledtensorrt-inference: AlexNet - INT8 - 4 - Disabledtensorrt-inference: ResNet50 - INT8 - 1 - Disabledtensorrt-inference: AlexNet - INT8 - 8 - Disabledmbw: Memory Copy, Fixed Block Size - 128 MiBtensorrt-inference: AlexNet - FP16 - 32 - Disabledmbw: Memory Copy, Fixed Block Size - 512 MiBtensorrt-inference: GoogleNet - FP16 - 8 - Disabledt-test1: 1t-test1: 2ramspeed: Add - Integertensorrt-inference: GoogleNet - FP16 - 1 - Disabledtensorrt-inference: ResNet152 - FP16 - 16 - Disabledtensorrt-inference: ResNet152 - FP16 - 32 - Disabledtensorrt-inference: GoogleNet - INT8 - 16 - Disabledtensorrt-inference: GoogleNet - INT8 - 32 - Disabledtensorrt-inference: GoogleNet - FP16 - 16 - Disabledtensorrt-inference: GoogleNet - FP16 - 32 - Disabledtensorrt-inference: ResNet50 - INT8 - 16 - Disabledtensorrt-inference: ResNet50 - INT8 - 32 - Disabledtensorrt-inference: ResNet50 - FP16 - 16 - Disabledtensorrt-inference: ResNet50 - FP16 - 32 - Disabledtensorrt-inference: ResNet152 - INT8 - 1 - Disabledtensorrt-inference: ResNet152 - FP16 - 8 - Disabledtensorrt-inference: ResNet152 - FP16 - 4 - Disabledramspeed: Copy - Integerramspeed: Triad - Integertensorrt-inference: GoogleNet - FP16 - 4 - Disabledramspeed: Scale - Integerramspeed: Average - Integercuda-mini-nbody: Cache Blockingglmark2: 1024 x 768glmark2: 1280 x 1024glmark2: 1920 x 1080lczero: BLAStensorrt-inference: ResNet152 - FP16 - 1 - Disabledlczero: CUDA + cuDNNj2dbench: Text Renderingj2dbench: Image Renderingj2dbench: Vector Graphics Renderingcompress-7zip: Compress Speed Testtensorrt-inference: GoogleNet - INT8 - 8 - Disabledglmark2: 800 x 600compress-zstd: Compressing ubuntu-16.04.3-server-i386.img, Compression Level 19cuda-mini-nbody: Originalbuild-linux-kernel: Time To Compilex264: H.264 Video EncodingJetson NanoNanonanoNano 5WNano 10w3.663.668.9335.8744.43342022.163439128.5547.83113.7610.2916942.0520.5911513411.6154.868.7014.6014.1827.3740.4840.6582.3014.6192.543450202344985.9280.3127.35794465.6116.9817.2852.1955.4793.3398.7523.8225.0144.4946.265.4516.4215.789544485685.12914278408.47136290464615.3410.091396226.12897658.51486283.59405049.181915127.284.0923795.122.442.446.005.673.072.442.446.0033.25296316.62297789.8837.6883.129.1512230.9016.211071118.8357.107.6410.9910.9225.7038.4530.1469.0613.7971.922963149296767.95110.7037.54655059.3912.1312.3039.6040.1269.6870.5717.3117.6532.1432.795.3211.7911.398048516866.29709067105.6710.12203238.703.0654792.313.643.648.9248.32354323.763542130.9853.68119.5611.9217944.0523.4114115311.9068.6010.0514.8014.6836.4145.8942.8992.1019.50102.303537213354296.2278.7226.90807982.8517.2517.4857.1257.9398.5099.9824.8325.3145.9146.647.6316.7716.279697529894.18933180608.4414.35412955.784.1721635.43OpenBenchmarking.org

CUDA Mini-Nbody

Test: SOA Data Layout

OpenBenchmarking.org(NBody^2)/s, More Is BetterCUDA Mini-Nbody 2015-11-10Test: SOA Data LayoutJetson NanonanoNano 5WNano 10w0.82351.6472.47053.2944.1175SE +/- 0.00, N = 3SE +/- 0.00, N = 3SE +/- 0.00, N = 3SE +/- 0.00, N = 33.662.442.443.64

CUDA Mini-Nbody

Test: Flush Denormals To Zero

OpenBenchmarking.org(NBody^2)/s, More Is BetterCUDA Mini-Nbody 2015-11-10Test: Flush Denormals To ZeroJetson NanonanoNano 5WNano 10w0.82351.6472.47053.2944.1175SE +/- 0.00, N = 3SE +/- 0.00, N = 3SE +/- 0.00, N = 3SE +/- 0.00, N = 33.662.442.443.64

CUDA Mini-Nbody

Test: Loop Unrolling

OpenBenchmarking.org(NBody^2)/s, More Is BetterCUDA Mini-Nbody 2015-11-10Test: Loop UnrollingJetson NanonanoNano 5WNano 10w246810SE +/- 0.03, N = 3SE +/- 0.00, N = 3SE +/- 0.00, N = 3SE +/- 0.00, N = 38.936.006.008.92

NVIDIA TensorRT Inference

Neural Network: GoogleNet - Precision: INT8 - Batch Size: 1 - DLA Cores: Disabled

OpenBenchmarking.orgImages Per Second, More Is BetterNVIDIA TensorRT InferenceNeural Network: GoogleNet - Precision: INT8 - Batch Size: 1 - DLA Cores: DisabledJetson NanoNano 5WNano 10w1122334455SE +/- 0.50, N = 3SE +/- 0.21, N = 3SE +/- 0.11, N = 335.8733.2548.32

XZ Compression

Compressing ubuntu-16.04.3-server-i386.img, Compression Level 9

OpenBenchmarking.orgSeconds, Fewer Is BetterXZ Compression 5.2.4Compressing ubuntu-16.04.3-server-i386.img, Compression Level 9Jetson Nano1020304050SE +/- 0.86, N = 344.431. (CC) gcc options: -pthread -fvisibility=hidden -O2

MBW

Test: Memory Copy - Array Size: 128 MiB

OpenBenchmarking.orgMiB/s, More Is BetterMBW 2018-09-08Test: Memory Copy - Array Size: 128 MiBJetson NanoNano 5WNano 10w8001600240032004000SE +/- 7.55, N = 3SE +/- 7.90, N = 3SE +/- 5.27, N = 33420296335431. (CC) gcc options: -O3 -march=native

NVIDIA TensorRT Inference

Neural Network: ResNet50 - Precision: INT8 - Batch Size: 8 - DLA Cores: Disabled

OpenBenchmarking.orgImages Per Second, More Is BetterNVIDIA TensorRT InferenceNeural Network: ResNet50 - Precision: INT8 - Batch Size: 8 - DLA Cores: DisabledJetson NanoNano 5WNano 10w612182430SE +/- 0.15, N = 3SE +/- 0.00, N = 3SE +/- 0.17, N = 322.1616.6223.76

MBW

Test: Memory Copy - Array Size: 512 MiB

OpenBenchmarking.orgMiB/s, More Is BetterMBW 2018-09-08Test: Memory Copy - Array Size: 512 MiBJetson NanoNano 5WNano 10w8001600240032004000SE +/- 12.45, N = 3SE +/- 3.31, N = 3SE +/- 2.80, N = 33439297735421. (CC) gcc options: -O3 -march=native

NVIDIA TensorRT Inference

Neural Network: AlexNet - Precision: INT8 - Batch Size: 32 - DLA Cores: Disabled

OpenBenchmarking.orgImages Per Second, More Is BetterNVIDIA TensorRT InferenceNeural Network: AlexNet - Precision: INT8 - Batch Size: 32 - DLA Cores: DisabledJetson NanoNano 5WNano 10w306090120150SE +/- 0.58, N = 3SE +/- 0.04, N = 3SE +/- 0.07, N = 3128.5589.88130.98

NVIDIA TensorRT Inference

Neural Network: GoogleNet - Precision: INT8 - Batch Size: 4 - DLA Cores: Disabled

OpenBenchmarking.orgImages Per Second, More Is BetterNVIDIA TensorRT InferenceNeural Network: GoogleNet - Precision: INT8 - Batch Size: 4 - DLA Cores: DisabledJetson NanoNano 5WNano 10w1224364860SE +/- 0.39, N = 3SE +/- 0.05, N = 3SE +/- 0.05, N = 347.8337.6853.68

NVIDIA TensorRT Inference

Neural Network: AlexNet - Precision: INT8 - Batch Size: 16 - DLA Cores: Disabled

OpenBenchmarking.orgImages Per Second, More Is BetterNVIDIA TensorRT InferenceNeural Network: AlexNet - Precision: INT8 - Batch Size: 16 - DLA Cores: DisabledJetson NanoNano 5WNano 10w306090120150SE +/- 1.48, N = 3SE +/- 0.09, N = 3SE +/- 0.17, N = 3113.7683.12119.56

NVIDIA TensorRT Inference

Neural Network: VGG16 - Precision: FP16 - Batch Size: 1 - DLA Cores: Disabled

OpenBenchmarking.orgImages Per Second, More Is BetterNVIDIA TensorRT InferenceNeural Network: VGG16 - Precision: FP16 - Batch Size: 1 - DLA Cores: DisabledJetson NanoNano 5WNano 10w3691215SE +/- 0.13, N = 8SE +/- 0.03, N = 3SE +/- 0.03, N = 310.299.1511.92

NVIDIA TensorRT Inference

Neural Network: AlexNet - Precision: FP16 - Batch Size: 16 - DLA Cores: Disabled

OpenBenchmarking.orgImages Per Second, More Is BetterNVIDIA TensorRT InferenceNeural Network: AlexNet - Precision: FP16 - Batch Size: 16 - DLA Cores: DisabledJetson NanoNano 5WNano 10w4080120160200SE +/- 1.25, N = 3SE +/- 1.89, N = 5SE +/- 0.31, N = 3169122179

NVIDIA TensorRT Inference

Neural Network: ResNet50 - Precision: FP16 - Batch Size: 8 - DLA Cores: Disabled

OpenBenchmarking.orgImages Per Second, More Is BetterNVIDIA TensorRT InferenceNeural Network: ResNet50 - Precision: FP16 - Batch Size: 8 - DLA Cores: DisabledJetson NanoNano 5WNano 10w1020304050SE +/- 0.20, N = 3SE +/- 0.02, N = 3SE +/- 0.03, N = 342.0530.9044.05

NVIDIA TensorRT Inference

Neural Network: ResNet50 - Precision: INT8 - Batch Size: 4 - DLA Cores: Disabled

OpenBenchmarking.orgImages Per Second, More Is BetterNVIDIA TensorRT InferenceNeural Network: ResNet50 - Precision: INT8 - Batch Size: 4 - DLA Cores: DisabledJetson NanoNano 5WNano 10w612182430SE +/- 0.30, N = 3SE +/- 0.05, N = 3SE +/- 0.01, N = 320.5916.2123.41

NVIDIA TensorRT Inference

Neural Network: AlexNet - Precision: FP16 - Batch Size: 4 - DLA Cores: Disabled

OpenBenchmarking.orgImages Per Second, More Is BetterNVIDIA TensorRT InferenceNeural Network: AlexNet - Precision: FP16 - Batch Size: 4 - DLA Cores: DisabledJetson NanoNano 5WNano 10w306090120150SE +/- 2.17, N = 12SE +/- 0.39, N = 3SE +/- 0.15, N = 3115107141

NVIDIA TensorRT Inference

Neural Network: AlexNet - Precision: FP16 - Batch Size: 8 - DLA Cores: Disabled

OpenBenchmarking.orgImages Per Second, More Is BetterNVIDIA TensorRT InferenceNeural Network: AlexNet - Precision: FP16 - Batch Size: 8 - DLA Cores: DisabledJetson NanoNano 5WNano 10w306090120150SE +/- 0.87, N = 3SE +/- 2.21, N = 3SE +/- 0.10, N = 3134111153

NVIDIA TensorRT Inference

Neural Network: VGG19 - Precision: FP16 - Batch Size: 4 - DLA Cores: Disabled

OpenBenchmarking.orgImages Per Second, More Is BetterNVIDIA TensorRT InferenceNeural Network: VGG19 - Precision: FP16 - Batch Size: 4 - DLA Cores: DisabledJetson NanoNano 5WNano 10w3691215SE +/- 0.08, N = 3SE +/- 0.01, N = 3SE +/- 0.02, N = 311.618.8311.90

NVIDIA TensorRT Inference

Neural Network: AlexNet - Precision: FP16 - Batch Size: 1 - DLA Cores: Disabled

OpenBenchmarking.orgImages Per Second, More Is BetterNVIDIA TensorRT InferenceNeural Network: AlexNet - Precision: FP16 - Batch Size: 1 - DLA Cores: DisabledJetson NanoNano 5WNano 10w1530456075SE +/- 1.49, N = 9SE +/- 0.16, N = 3SE +/- 0.66, N = 354.8657.1068.60

NVIDIA TensorRT Inference

Neural Network: VGG19 - Precision: FP16 - Batch Size: 1 - DLA Cores: Disabled

OpenBenchmarking.orgImages Per Second, More Is BetterNVIDIA TensorRT InferenceNeural Network: VGG19 - Precision: FP16 - Batch Size: 1 - DLA Cores: DisabledJetson NanoNano 5WNano 10w3691215SE +/- 0.09, N = 3SE +/- 0.00, N = 3SE +/- 0.03, N = 38.707.6410.05

NVIDIA TensorRT Inference

Neural Network: VGG16 - Precision: FP16 - Batch Size: 8 - DLA Cores: Disabled

OpenBenchmarking.orgImages Per Second, More Is BetterNVIDIA TensorRT InferenceNeural Network: VGG16 - Precision: FP16 - Batch Size: 8 - DLA Cores: DisabledJetson NanoNano 5WNano 10w48121620SE +/- 0.00, N = 3SE +/- 0.01, N = 3SE +/- 0.02, N = 314.6010.9914.80

NVIDIA TensorRT Inference

Neural Network: VGG16 - Precision: FP16 - Batch Size: 4 - DLA Cores: Disabled

OpenBenchmarking.orgImages Per Second, More Is BetterNVIDIA TensorRT InferenceNeural Network: VGG16 - Precision: FP16 - Batch Size: 4 - DLA Cores: DisabledJetson NanoNano 5WNano 10w48121620SE +/- 0.08, N = 3SE +/- 0.00, N = 3SE +/- 0.01, N = 314.1810.9214.68

NVIDIA TensorRT Inference

Neural Network: ResNet50 - Precision: FP16 - Batch Size: 1 - DLA Cores: Disabled

OpenBenchmarking.orgImages Per Second, More Is BetterNVIDIA TensorRT InferenceNeural Network: ResNet50 - Precision: FP16 - Batch Size: 1 - DLA Cores: DisabledJetson NanoNano 5WNano 10w816243240SE +/- 0.34, N = 9SE +/- 0.11, N = 3SE +/- 0.12, N = 327.3725.7036.41

NVIDIA TensorRT Inference

Neural Network: AlexNet - Precision: INT8 - Batch Size: 1 - DLA Cores: Disabled

OpenBenchmarking.orgImages Per Second, More Is BetterNVIDIA TensorRT InferenceNeural Network: AlexNet - Precision: INT8 - Batch Size: 1 - DLA Cores: DisabledJetson NanoNano 5WNano 10w1020304050SE +/- 0.71, N = 3SE +/- 0.17, N = 3SE +/- 0.14, N = 340.4838.4545.89

NVIDIA TensorRT Inference

Neural Network: ResNet50 - Precision: FP16 - Batch Size: 4 - DLA Cores: Disabled

OpenBenchmarking.orgImages Per Second, More Is BetterNVIDIA TensorRT InferenceNeural Network: ResNet50 - Precision: FP16 - Batch Size: 4 - DLA Cores: DisabledJetson NanoNano 5WNano 10w1020304050SE +/- 0.26, N = 3SE +/- 0.01, N = 3SE +/- 0.02, N = 340.6530.1442.89

NVIDIA TensorRT Inference

Neural Network: AlexNet - Precision: INT8 - Batch Size: 4 - DLA Cores: Disabled

OpenBenchmarking.orgImages Per Second, More Is BetterNVIDIA TensorRT InferenceNeural Network: AlexNet - Precision: INT8 - Batch Size: 4 - DLA Cores: DisabledJetson NanoNano 5WNano 10w20406080100SE +/- 1.37, N = 4SE +/- 0.01, N = 3SE +/- 0.18, N = 382.3069.0692.10

NVIDIA TensorRT Inference

Neural Network: ResNet50 - Precision: INT8 - Batch Size: 1 - DLA Cores: Disabled

OpenBenchmarking.orgImages Per Second, More Is BetterNVIDIA TensorRT InferenceNeural Network: ResNet50 - Precision: INT8 - Batch Size: 1 - DLA Cores: DisabledJetson NanoNano 5WNano 10w510152025SE +/- 0.12, N = 3SE +/- 0.13, N = 3SE +/- 0.15, N = 314.6113.7919.50

NVIDIA TensorRT Inference

Neural Network: AlexNet - Precision: INT8 - Batch Size: 8 - DLA Cores: Disabled

OpenBenchmarking.orgImages Per Second, More Is BetterNVIDIA TensorRT InferenceNeural Network: AlexNet - Precision: INT8 - Batch Size: 8 - DLA Cores: DisabledJetson NanoNano 5WNano 10w20406080100SE +/- 0.96, N = 3SE +/- 0.03, N = 3SE +/- 0.32, N = 392.5471.92102.30

MBW

Test: Memory Copy, Fixed Block Size - Array Size: 128 MiB

OpenBenchmarking.orgMiB/s, More Is BetterMBW 2018-09-08Test: Memory Copy, Fixed Block Size - Array Size: 128 MiBJetson NanoNano 5WNano 10w8001600240032004000SE +/- 7.52, N = 3SE +/- 10.41, N = 3SE +/- 2.59, N = 33450296335371. (CC) gcc options: -O3 -march=native

NVIDIA TensorRT Inference

Neural Network: AlexNet - Precision: FP16 - Batch Size: 32 - DLA Cores: Disabled

OpenBenchmarking.orgImages Per Second, More Is BetterNVIDIA TensorRT InferenceNeural Network: AlexNet - Precision: FP16 - Batch Size: 32 - DLA Cores: DisabledJetson NanoNano 5WNano 10w50100150200250SE +/- 0.75, N = 3SE +/- 0.07, N = 3SE +/- 0.21, N = 3202149213

MBW

Test: Memory Copy, Fixed Block Size - Array Size: 512 MiB

OpenBenchmarking.orgMiB/s, More Is BetterMBW 2018-09-08Test: Memory Copy, Fixed Block Size - Array Size: 512 MiBJetson NanoNano 5WNano 10w8001600240032004000SE +/- 14.16, N = 3SE +/- 1.52, N = 3SE +/- 1.99, N = 33449296735421. (CC) gcc options: -O3 -march=native

NVIDIA TensorRT Inference

Neural Network: GoogleNet - Precision: FP16 - Batch Size: 8 - DLA Cores: Disabled

OpenBenchmarking.orgImages Per Second, More Is BetterNVIDIA TensorRT InferenceNeural Network: GoogleNet - Precision: FP16 - Batch Size: 8 - DLA Cores: DisabledJetson NanoNano 5WNano 10w20406080100SE +/- 0.10, N = 3SE +/- 0.03, N = 3SE +/- 0.08, N = 385.9267.9596.22

t-test1

Threads: 1

OpenBenchmarking.orgSeconds, Fewer Is Bettert-test1 2017-01-13Threads: 1Jetson NanoNano 5WNano 10w20406080100SE +/- 0.23, N = 3SE +/- 0.05, N = 3SE +/- 0.32, N = 380.31110.7078.721. (CC) gcc options: -pthread

t-test1

Threads: 2

OpenBenchmarking.orgSeconds, Fewer Is Bettert-test1 2017-01-13Threads: 2Jetson NanoNano 5WNano 10w918273645SE +/- 0.07, N = 3SE +/- 0.10, N = 3SE +/- 0.04, N = 327.3537.5426.901. (CC) gcc options: -pthread

RAMspeed SMP

Type: Add - Benchmark: Integer

OpenBenchmarking.orgMB/s, More Is BetterRAMspeed SMP 3.5.0Type: Add - Benchmark: IntegerJetson NanoNano 5WNano 10w2K4K6K8K10K7944655080791. (CC) gcc options: -O3 -march=native

NVIDIA TensorRT Inference

Neural Network: GoogleNet - Precision: FP16 - Batch Size: 1 - DLA Cores: Disabled

OpenBenchmarking.orgImages Per Second, More Is BetterNVIDIA TensorRT InferenceNeural Network: GoogleNet - Precision: FP16 - Batch Size: 1 - DLA Cores: DisabledJetson NanoNano 5WNano 10w20406080100SE +/- 0.96, N = 9SE +/- 0.02, N = 3SE +/- 0.25, N = 365.6159.3982.85

NVIDIA TensorRT Inference

Neural Network: ResNet152 - Precision: FP16 - Batch Size: 16 - DLA Cores: Disabled

OpenBenchmarking.orgImages Per Second, More Is BetterNVIDIA TensorRT InferenceNeural Network: ResNet152 - Precision: FP16 - Batch Size: 16 - DLA Cores: DisabledJetson NanoNano 5WNano 10w48121620SE +/- 0.04, N = 3SE +/- 0.01, N = 3SE +/- 0.02, N = 316.9812.1317.25

NVIDIA TensorRT Inference

Neural Network: ResNet152 - Precision: FP16 - Batch Size: 32 - DLA Cores: Disabled

OpenBenchmarking.orgImages Per Second, More Is BetterNVIDIA TensorRT InferenceNeural Network: ResNet152 - Precision: FP16 - Batch Size: 32 - DLA Cores: DisabledJetson NanoNano 5WNano 10w48121620SE +/- 0.01, N = 3SE +/- 0.01, N = 3SE +/- 0.01, N = 317.2812.3017.48

NVIDIA TensorRT Inference

Neural Network: GoogleNet - Precision: INT8 - Batch Size: 16 - DLA Cores: Disabled

OpenBenchmarking.orgImages Per Second, More Is BetterNVIDIA TensorRT InferenceNeural Network: GoogleNet - Precision: INT8 - Batch Size: 16 - DLA Cores: DisabledJetson NanoNano 5WNano 10w1326395265SE +/- 0.34, N = 3SE +/- 0.02, N = 3SE +/- 0.04, N = 352.1939.6057.12

NVIDIA TensorRT Inference

Neural Network: GoogleNet - Precision: INT8 - Batch Size: 32 - DLA Cores: Disabled

OpenBenchmarking.orgImages Per Second, More Is BetterNVIDIA TensorRT InferenceNeural Network: GoogleNet - Precision: INT8 - Batch Size: 32 - DLA Cores: DisabledJetson NanoNano 5WNano 10w1326395265SE +/- 0.21, N = 3SE +/- 0.00, N = 3SE +/- 0.01, N = 355.4740.1257.93

NVIDIA TensorRT Inference

Neural Network: GoogleNet - Precision: FP16 - Batch Size: 16 - DLA Cores: Disabled

OpenBenchmarking.orgImages Per Second, More Is BetterNVIDIA TensorRT InferenceNeural Network: GoogleNet - Precision: FP16 - Batch Size: 16 - DLA Cores: DisabledJetson NanoNano 5WNano 10w20406080100SE +/- 1.84, N = 3SE +/- 0.03, N = 3SE +/- 0.06, N = 393.3369.6898.50

NVIDIA TensorRT Inference

Neural Network: GoogleNet - Precision: FP16 - Batch Size: 32 - DLA Cores: Disabled

OpenBenchmarking.orgImages Per Second, More Is BetterNVIDIA TensorRT InferenceNeural Network: GoogleNet - Precision: FP16 - Batch Size: 32 - DLA Cores: DisabledJetson NanoNano 5WNano 10w20406080100SE +/- 0.20, N = 3SE +/- 0.05, N = 3SE +/- 0.04, N = 398.7570.5799.98

NVIDIA TensorRT Inference

Neural Network: ResNet50 - Precision: INT8 - Batch Size: 16 - DLA Cores: Disabled

OpenBenchmarking.orgImages Per Second, More Is BetterNVIDIA TensorRT InferenceNeural Network: ResNet50 - Precision: INT8 - Batch Size: 16 - DLA Cores: DisabledJetson NanoNano 5WNano 10w612182430SE +/- 0.03, N = 3SE +/- 0.01, N = 3SE +/- 0.01, N = 323.8217.3124.83

NVIDIA TensorRT Inference

Neural Network: ResNet50 - Precision: INT8 - Batch Size: 32 - DLA Cores: Disabled

OpenBenchmarking.orgImages Per Second, More Is BetterNVIDIA TensorRT InferenceNeural Network: ResNet50 - Precision: INT8 - Batch Size: 32 - DLA Cores: DisabledJetson NanoNano 5WNano 10w612182430SE +/- 0.06, N = 3SE +/- 0.00, N = 3SE +/- 0.00, N = 325.0117.6525.31

NVIDIA TensorRT Inference

Neural Network: ResNet50 - Precision: FP16 - Batch Size: 16 - DLA Cores: Disabled

OpenBenchmarking.orgImages Per Second, More Is BetterNVIDIA TensorRT InferenceNeural Network: ResNet50 - Precision: FP16 - Batch Size: 16 - DLA Cores: DisabledJetson NanoNano 5WNano 10w1020304050SE +/- 0.39, N = 3SE +/- 0.01, N = 3SE +/- 0.03, N = 344.4932.1445.91

NVIDIA TensorRT Inference

Neural Network: ResNet50 - Precision: FP16 - Batch Size: 32 - DLA Cores: Disabled

OpenBenchmarking.orgImages Per Second, More Is BetterNVIDIA TensorRT InferenceNeural Network: ResNet50 - Precision: FP16 - Batch Size: 32 - DLA Cores: DisabledJetson NanoNano 5WNano 10w1122334455SE +/- 0.01, N = 3SE +/- 0.01, N = 3SE +/- 0.04, N = 346.2632.7946.64

NVIDIA TensorRT Inference

Neural Network: ResNet152 - Precision: INT8 - Batch Size: 1 - DLA Cores: Disabled

OpenBenchmarking.orgImages Per Second, More Is BetterNVIDIA TensorRT InferenceNeural Network: ResNet152 - Precision: INT8 - Batch Size: 1 - DLA Cores: DisabledJetson NanoNano 5WNano 10w246810SE +/- 0.01, N = 3SE +/- 0.01, N = 3SE +/- 0.02, N = 35.455.327.63

NVIDIA TensorRT Inference

Neural Network: ResNet152 - Precision: FP16 - Batch Size: 8 - DLA Cores: Disabled

OpenBenchmarking.orgImages Per Second, More Is BetterNVIDIA TensorRT InferenceNeural Network: ResNet152 - Precision: FP16 - Batch Size: 8 - DLA Cores: DisabledJetson NanoNano 5WNano 10w48121620SE +/- 0.08, N = 3SE +/- 0.00, N = 3SE +/- 0.01, N = 316.4211.7916.77

NVIDIA TensorRT Inference

Neural Network: ResNet152 - Precision: FP16 - Batch Size: 4 - DLA Cores: Disabled

OpenBenchmarking.orgImages Per Second, More Is BetterNVIDIA TensorRT InferenceNeural Network: ResNet152 - Precision: FP16 - Batch Size: 4 - DLA Cores: DisabledJetson NanoNano 5WNano 10w48121620SE +/- 0.07, N = 3SE +/- 0.01, N = 3SE +/- 0.01, N = 315.7811.3916.27

RAMspeed SMP

Type: Copy - Benchmark: Integer

OpenBenchmarking.orgMB/s, More Is BetterRAMspeed SMP 3.5.0Type: Copy - Benchmark: IntegerJetson NanoNano 5WNano 10w2K4K6K8K10K9544804896971. (CC) gcc options: -O3 -march=native

RAMspeed SMP

Type: Triad - Benchmark: Integer

OpenBenchmarking.orgMB/s, More Is BetterRAMspeed SMP 3.5.0Type: Triad - Benchmark: IntegerJetson NanoNano 5WNano 10w110022003300440055004856516852981. (CC) gcc options: -O3 -march=native

NVIDIA TensorRT Inference

Neural Network: GoogleNet - Precision: FP16 - Batch Size: 4 - DLA Cores: Disabled

OpenBenchmarking.orgImages Per Second, More Is BetterNVIDIA TensorRT InferenceNeural Network: GoogleNet - Precision: FP16 - Batch Size: 4 - DLA Cores: DisabledJetson NanoNano 5WNano 10w20406080100SE +/- 1.10, N = 12SE +/- 0.02, N = 3SE +/- 0.10, N = 385.1266.2994.18

RAMspeed SMP

Type: Scale - Benchmark: Integer

OpenBenchmarking.orgMB/s, More Is BetterRAMspeed SMP 3.5.0Type: Scale - Benchmark: IntegerJetson NanoNano 5WNano 10w2K4K6K8K10K9142709093311. (CC) gcc options: -O3 -march=native

RAMspeed SMP

Type: Average - Benchmark: Integer

OpenBenchmarking.orgMB/s, More Is BetterRAMspeed SMP 3.5.0Type: Average - Benchmark: IntegerJetson NanoNano 5WNano 10w2K4K6K8K10K7840671080601. (CC) gcc options: -O3 -march=native

CUDA Mini-Nbody

Test: Cache Blocking

OpenBenchmarking.org(NBody^2)/s, More Is BetterCUDA Mini-Nbody 2015-11-10Test: Cache BlockingJetson NanonanoNano 5WNano 10w246810SE +/- 0.00, N = 3SE +/- 0.00, N = 3SE +/- 0.00, N = 3SE +/- 0.00, N = 38.475.675.678.44

GLmark2

Resolution: 1024 x 768

OpenBenchmarking.orgScore, More Is BetterGLmark2 276Resolution: 1024 x 768Jetson Nano300600900120015001362

GLmark2

Resolution: 1280 x 1024

OpenBenchmarking.orgScore, More Is BetterGLmark2 276Resolution: 1280 x 1024Jetson Nano2004006008001000904

GLmark2

Resolution: 1920 x 1080

OpenBenchmarking.orgScore, More Is BetterGLmark2 276Resolution: 1920 x 1080Jetson Nano140280420560700646

LeelaChessZero

Backend: BLAS

OpenBenchmarking.orgNodes Per Second, More Is BetterLeelaChessZero 0.20.1Backend: BLASJetson Nano48121620SE +/- 0.10, N = 315.341. (CXX) g++ options: -lpthread -lz

NVIDIA TensorRT Inference

Neural Network: ResNet152 - Precision: FP16 - Batch Size: 1 - DLA Cores: Disabled

OpenBenchmarking.orgImages Per Second, More Is BetterNVIDIA TensorRT InferenceNeural Network: ResNet152 - Precision: FP16 - Batch Size: 1 - DLA Cores: DisabledJetson NanoNano 5WNano 10w48121620SE +/- 0.05, N = 3SE +/- 0.00, N = 3SE +/- 0.04, N = 310.0910.1214.35

LeelaChessZero

Backend: CUDA + cuDNN

OpenBenchmarking.orgNodes Per Second, More Is BetterLeelaChessZero 0.20.1Backend: CUDA + cuDNNJetson Nano306090120150SE +/- 0.64, N = 31391. (CXX) g++ options: -lpthread -lz

Java 2D Microbenchmark

Rendering Test: Text Rendering

OpenBenchmarking.orgUnits Per Second, More Is BetterJava 2D Microbenchmark 1.0Rendering Test: Text RenderingJetson Nano13002600390052006500SE +/- 34.48, N = 46226.12

Java 2D Microbenchmark

Rendering Test: Image Rendering

OpenBenchmarking.orgUnits Per Second, More Is BetterJava 2D Microbenchmark 1.0Rendering Test: Image RenderingJetson Nano200K400K600K800K1000KSE +/- 1827.16, N = 4897658.51

Java 2D Microbenchmark

Rendering Test: Vector Graphics Rendering

OpenBenchmarking.orgUnits Per Second, More Is BetterJava 2D Microbenchmark 1.0Rendering Test: Vector Graphics RenderingJetson Nano100K200K300K400K500KSE +/- 983.45, N = 4486283.59

7-Zip Compression

Compress Speed Test

OpenBenchmarking.orgMIPS, More Is Better7-Zip Compression 16.02Compress Speed TestJetson NanoNano 5WNano 10w9001800270036004500SE +/- 17.21, N = 3SE +/- 7.09, N = 3SE +/- 2.96, N = 34050203241291. (CXX) g++ options: -pipe -lpthread

NVIDIA TensorRT Inference

Neural Network: GoogleNet - Precision: INT8 - Batch Size: 8 - DLA Cores: Disabled

OpenBenchmarking.orgImages Per Second, More Is BetterNVIDIA TensorRT InferenceNeural Network: GoogleNet - Precision: INT8 - Batch Size: 8 - DLA Cores: DisabledJetson NanoNano 5WNano 10w1326395265SE +/- 0.47, N = 3SE +/- 0.02, N = 3SE +/- 0.03, N = 349.1838.7055.78

GLmark2

Resolution: 800 x 600

OpenBenchmarking.orgScore, More Is BetterGLmark2 276Resolution: 800 x 600Jetson Nano4008001200160020001915

Zstd Compression

Compressing ubuntu-16.04.3-server-i386.img, Compression Level 19

OpenBenchmarking.orgSeconds, Fewer Is BetterZstd Compression 1.3.4Compressing ubuntu-16.04.3-server-i386.img, Compression Level 19Jetson Nano306090120150SE +/- 0.22, N = 3127.281. (CC) gcc options: -O3 -pthread -lz -llzma

CUDA Mini-Nbody

Test: Original

OpenBenchmarking.org(NBody^2)/s, More Is BetterCUDA Mini-Nbody 2015-11-10Test: OriginalJetson NanonanoNano 5WNano 10w0.93831.87662.81493.75324.6915SE +/- 0.01, N = 3SE +/- 0.00, N = 3SE +/- 0.01, N = 3SE +/- 0.00, N = 34.093.073.064.17

Timed Linux Kernel Compilation

Time To Compile

OpenBenchmarking.orgSeconds, Fewer Is BetterTimed Linux Kernel Compilation 4.18Time To CompileJetson NanoNano 5WNano 10w12002400360048006000SE +/- 13.46, N = 3SE +/- 2.37, N = 3SE +/- 2.72, N = 3237954792163

x264

H.264 Video Encoding

OpenBenchmarking.orgFrames Per Second, More Is Betterx264 2018-09-25H.264 Video EncodingJetson NanoNano 5WNano 10w1.22182.44363.66544.88726.109SE +/- 0.08, N = 3SE +/- 0.03, N = 3SE +/- 0.08, N = 45.122.315.431. (CC) gcc options: -ldl -lm -lpthread

Phoronix Test Suite v10.8.4