ARMv8 rev 1 testing with a jetson_tx1 and NVIDIA Tegra X1 on Ubuntu 16.04 via the Phoronix Test Suite.
Compare your own system(s) to this result file with the
Phoronix Test Suite by running the command:
phoronix-test-suite benchmark 1903178-SP-1903167SP11 Jetson Prep - Phoronix Test Suite Jetson Prep ARMv8 rev 1 testing with a jetson_tx1 and NVIDIA Tegra X1 on Ubuntu 16.04 via the Phoronix Test Suite.
HTML result view exported from: https://openbenchmarking.org/result/1903178-SP-1903167SP11&rdt&grs .
Jetson Prep Processor Motherboard Memory Disk Graphics Monitor OS Kernel Desktop Display Server Display Driver OpenGL Vulkan Compiler File-System Screen Resolution Jetson AGX Xavier Jetson TX2 Max-P Jetson TX2 Max-Q Raspberry Pi 3 Model B+ ASUS TinkerBoard Jetson TX1 Max-P ARMv8 rev 0 @ 2.27GHz (8 Cores) jetson-xavier 16384MB 31GB HBG4a2 NVIDIA Tegra Xavier VE228 Ubuntu 18.04 4.9.108-tegra (aarch64) Unity 7.5.0 X Server 1.19.6 NVIDIA 31.0.2 4.6.0 1.1.76 GCC 7.3.0 + CUDA 10.0 ext4 1920x1080 ARMv8 rev 3 @ 2.04GHz (4 Cores / 6 Threads) quill 8192MB 31GB 032G34 NVIDIA TEGRA Ubuntu 16.04 4.4.38-tegra (aarch64) Unity 7.4.0 X Server 1.18.4 NVIDIA 28.2.1 4.5.0 GCC 5.4.0 20160609 + CUDA 9.0 ARMv8 rev 3 @ 1.27GHz (4 Cores / 6 Threads) ARMv7 rev 4 @ 1.40GHz (4 Cores) BCM2835 Raspberry Pi 3 Model B Plus Rev 1.3 926MB 32GB GB2MW BCM2708 Raspbian 9.6 4.19.23-v7+ (armv7l) LXDE X Server 1.19.2 GCC 6.3.0 20170516 656x416 ARMv7 rev 1 @ 1.80GHz (4 Cores) Rockchip (Device Tree) 2048MB 32GB GB1QT Debian 9.0 4.4.16-00006-g4431f98-dirty (armv7l) X Server 1.18.4 1024x768 ARMv8 rev 1 @ 1.73GHz (4 Cores) jetson_tx1 4096MB 16GB 016G32 NVIDIA Tegra X1 VE228 Ubuntu 16.04 4.4.38-tegra (aarch64) Unity 7.4.5 NVIDIA 28.1.0 4.5.0 1.0.8 GCC 5.4.0 20160609 1920x1080 OpenBenchmarking.org Compiler Details - Jetson AGX Xavier: --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 - Jetson TX2 Max-P: --build=aarch64-linux-gnu --disable-browser-plugin --disable-libquadmath --disable-werror --enable-checking=release --enable-clocale=gnu --enable-fix-cortex-a53-843419 --enable-gnu-unique-object --enable-gtk-cairo --enable-java-awt=gtk --enable-java-home --enable-languages=c,ada,c++,java,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 --target=aarch64-linux-gnu --with-arch-directory=aarch64 --with-default-libstdcxx-abi=new -v - Jetson TX2 Max-Q: --build=aarch64-linux-gnu --disable-browser-plugin --disable-libquadmath --disable-werror --enable-checking=release --enable-clocale=gnu --enable-fix-cortex-a53-843419 --enable-gnu-unique-object --enable-gtk-cairo --enable-java-awt=gtk --enable-java-home --enable-languages=c,ada,c++,java,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 --target=aarch64-linux-gnu --with-arch-directory=aarch64 --with-default-libstdcxx-abi=new -v - Raspberry Pi 3 Model B+: --build=arm-linux-gnueabihf --disable-browser-plugin --disable-libitm --disable-libquadmath --disable-sjlj-exceptions --enable-checking=release --enable-clocale=gnu --enable-gnu-unique-object --enable-gtk-cairo --enable-java-awt=gtk --enable-java-home --enable-languages=c,ada,c++,java,go,d,fortran,objc,obj-c++ --enable-libstdcxx-debug --enable-libstdcxx-time=yes --enable-multiarch --enable-nls --enable-objc-gc=auto --enable-plugin --enable-shared --enable-threads=posix --host=arm-linux-gnueabihf --program-prefix=arm-linux-gnueabihf- --target=arm-linux-gnueabihf --with-arch-directory=arm --with-arch=armv6 --with-default-libstdcxx-abi=new --with-float=hard --with-fpu=vfp --with-target-system-zlib -v - ASUS TinkerBoard: --build=arm-linux-gnueabihf --disable-browser-plugin --disable-libitm --disable-libquadmath --disable-sjlj-exceptions --enable-checking=release --enable-clocale=gnu --enable-default-pie --enable-gnu-unique-object --enable-gtk-cairo --enable-java-awt=gtk --enable-java-home --enable-languages=c,ada,c++,java,go,d,fortran,objc,obj-c++ --enable-libstdcxx-debug --enable-libstdcxx-time=yes --enable-multiarch --enable-nls --enable-objc-gc=auto --enable-plugin --enable-shared --enable-threads=posix --host=arm-linux-gnueabihf --program-prefix=arm-linux-gnueabihf- --target=arm-linux-gnueabihf --with-arch-directory=arm --with-arch=armv7-a --with-default-libstdcxx-abi=new --with-float=hard --with-fpu=vfpv3-d16 --with-mode=thumb --with-target-system-zlib -v - Jetson TX1 Max-P: --build=aarch64-linux-gnu --disable-browser-plugin --disable-libquadmath --disable-werror --enable-checking=release --enable-clocale=gnu --enable-fix-cortex-a53-843419 --enable-gnu-unique-object --enable-gtk-cairo --enable-java-awt=gtk --enable-java-home --enable-languages=c,ada,c++,java,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 --target=aarch64-linux-gnu --with-arch-directory=aarch64 --with-default-libstdcxx-abi=new -v Processor Details - Jetson AGX Xavier: Scaling Governor: tegra_cpufreq schedutil - Jetson TX2 Max-P: Scaling Governor: tegra_cpufreq schedutil - Jetson TX2 Max-Q: Scaling Governor: tegra_cpufreq schedutil - Raspberry Pi 3 Model B+: Scaling Governor: BCM2835 Freq ondemand - ASUS TinkerBoard: Scaling Governor: cpufreq-dt interactive - Jetson TX1 Max-P: Scaling Governor: tegra-cpufreq interactive Python Details - Jetson AGX Xavier: Python 2.7.15rc1 + Python 3.6.7 - Jetson TX2 Max-P: Python 2.7.12 + Python 3.5.2 - Jetson TX2 Max-Q: Python 2.7.12 + Python 3.5.2 - Raspberry Pi 3 Model B+: Python 2.7.13 + Python 3.5.3 - ASUS TinkerBoard: Python 2.7.13 + Python 3.5.3 - Jetson TX1 Max-P: Python 2.7.12 + Python 3.5.2
Jetson Prep compress-7zip: Compress Speed Test tensorrt-inference: VGG19 - FP16 - 32 - Disabled tensorrt-inference: VGG16 - FP16 - 32 - Disabled tensorrt-inference: ResNet152 - FP16 - 4 - Disabled tensorrt-inference: VGG19 - FP16 - 4 - Disabled tensorrt-inference: VGG16 - FP16 - 4 - Disabled tensorrt-inference: ResNet152 - FP16 - 32 - Disabled tensorrt-inference: AlexNet - INT8 - 4 - Disabled tensorrt-inference: ResNet50 - FP16 - 4 - Disabled ttsiod-renderer: Phong Rendering With Soft-Shadow Mapping tensorrt-inference: ResNet50 - FP16 - 32 - Disabled cuda-mini-nbody: Original pybench: Total For Average Test Times encode-flac: WAV To FLAC compress-zstd: Compressing ubuntu-16.04.3-server-i386.img, Compression Level 19 rust-prime: Prime Number Test To 200,000,000 c-ray: Total Time - 4K, 16 Rays Per Pixel tensorrt-inference: GoogleNet - FP16 - 32 - Disabled tensorrt-inference: AlexNet - FP16 - 32 - Disabled tensorrt-inference: GoogleNet - FP16 - 4 - Disabled tensorrt-inference: VGG19 - INT8 - 32 - Disabled tensorrt-inference: VGG16 - INT8 - 32 - Disabled tensorrt-inference: ResNet152 - INT8 - 32 - Disabled tensorrt-inference: ResNet50 - INT8 - 32 - Disabled tensorrt-inference: ResNet152 - INT8 - 4 - Disabled tensorrt-inference: VGG19 - INT8 - 4 - Disabled tensorrt-inference: ResNet50 - INT8 - 4 - Disabled tensorrt-inference: VGG16 - INT8 - 4 - Disabled opencv-bench: tensorrt-inference: GoogleNet - INT8 - 32 - Disabled tensorrt-inference: AlexNet - INT8 - 32 - Disabled tensorrt-inference: GoogleNet - INT8 - 4 - Disabled tesseract-ocr: Time To OCR 7 Images lczero: CUDA + cuDNN FP16 lczero: CUDA + cuDNN lczero: BLAS glmark2: 1920 x 1080 tensorrt-inference: AlexNet - FP16 - 4 - Disabled Jetson AGX Xavier Jetson TX2 Max-P Jetson TX2 Max-Q Raspberry Pi 3 Model B+ ASUS TinkerBoard Jetson TX1 Max-P 19212 203.96 247.95 224.19 172.50 208.76 259.82 1143 547.50 133 636 47.13 3007 54.47 80.06 32.37 355 1006 2038 796 394.66 475.08 493.22 1215.08 372.73 265.81 902.78 303.78 128 1693 3143 1146 71.94 2515.01 952.89 47.62 2876 1200 5593 29.83 36.87 35.11 26.56 32.64 41.91 184 92.28 49.26 111 8.24 5408 65.07 144.97 104.96 585 233 462 197 15.92 19.91 22.07 59.69 18.29 14.32 49.97 17.56 296 130 301 113 264 3294 23.94 29.83 27.34 21.04 25.99 32.67 148 72.01 28.85 86.08 6.77 8735 104.28 253.80 170.25 869 179 374 156 12.59 15.79 17.36 47.15 14.50 11.45 39.15 14.24 493 104 237 88.88 216 2013 17.66 20913 339.53 342.23 1097.69 2030 2.74 2836 21.22 11502 279.05 496.62 1821.05 1718 4508 45.09 6339 79.20 145.80 128.45 753 OpenBenchmarking.org
7-Zip Compression Compress Speed Test OpenBenchmarking.org MIPS, More Is Better 7-Zip Compression 16.02 Compress Speed Test Jetson AGX Xavier Jetson TX2 Max-P Jetson TX2 Max-Q Raspberry Pi 3 Model B+ ASUS TinkerBoard Jetson TX1 Max-P 4K 8K 12K 16K 20K SE +/- 274.18, N = 12 SE +/- 20.85, N = 3 SE +/- 13.05, N = 3 SE +/- 23.74, N = 11 SE +/- 34.93, N = 3 SE +/- 13.43, N = 3 19212 5593 3294 2013 2836 4508 1. (CXX) g++ options: -pipe -lpthread
NVIDIA TensorRT Inference Neural Network: VGG19 - Precision: FP16 - Batch Size: 32 - DLA Cores: Disabled OpenBenchmarking.org Images Per Second, More Is Better NVIDIA TensorRT Inference Neural Network: VGG19 - Precision: FP16 - Batch Size: 32 - DLA Cores: Disabled Jetson AGX Xavier Jetson TX2 Max-P Jetson TX2 Max-Q 40 80 120 160 200 SE +/- 0.04, N = 3 SE +/- 0.05, N = 3 SE +/- 0.07, N = 3 203.96 29.83 23.94
NVIDIA TensorRT Inference Neural Network: VGG16 - Precision: FP16 - Batch Size: 32 - DLA Cores: Disabled OpenBenchmarking.org Images Per Second, More Is Better NVIDIA TensorRT Inference Neural Network: VGG16 - Precision: FP16 - Batch Size: 32 - DLA Cores: Disabled Jetson AGX Xavier Jetson TX2 Max-P Jetson TX2 Max-Q 50 100 150 200 250 SE +/- 0.12, N = 3 SE +/- 0.31, N = 3 SE +/- 0.18, N = 3 247.95 36.87 29.83
NVIDIA TensorRT Inference Neural Network: ResNet152 - Precision: FP16 - Batch Size: 4 - DLA Cores: Disabled OpenBenchmarking.org Images Per Second, More Is Better NVIDIA TensorRT Inference Neural Network: ResNet152 - Precision: FP16 - Batch Size: 4 - DLA Cores: Disabled Jetson AGX Xavier Jetson TX2 Max-P Jetson TX2 Max-Q 50 100 150 200 250 SE +/- 0.22, N = 3 SE +/- 0.36, N = 3 SE +/- 0.34, N = 3 224.19 35.11 27.34
NVIDIA TensorRT Inference Neural Network: VGG19 - Precision: FP16 - Batch Size: 4 - DLA Cores: Disabled OpenBenchmarking.org Images Per Second, More Is Better NVIDIA TensorRT Inference Neural Network: VGG19 - Precision: FP16 - Batch Size: 4 - DLA Cores: Disabled Jetson AGX Xavier Jetson TX2 Max-P Jetson TX2 Max-Q 40 80 120 160 200 SE +/- 0.50, N = 3 SE +/- 0.38, N = 3 SE +/- 0.34, N = 3 172.50 26.56 21.04
NVIDIA TensorRT Inference Neural Network: VGG16 - Precision: FP16 - Batch Size: 4 - DLA Cores: Disabled OpenBenchmarking.org Images Per Second, More Is Better NVIDIA TensorRT Inference Neural Network: VGG16 - Precision: FP16 - Batch Size: 4 - DLA Cores: Disabled Jetson AGX Xavier Jetson TX2 Max-P Jetson TX2 Max-Q 50 100 150 200 250 SE +/- 0.10, N = 3 SE +/- 0.50, N = 4 SE +/- 0.13, N = 3 208.76 32.64 25.99
NVIDIA TensorRT Inference Neural Network: ResNet152 - Precision: FP16 - Batch Size: 32 - DLA Cores: Disabled OpenBenchmarking.org Images Per Second, More Is Better NVIDIA TensorRT Inference Neural Network: ResNet152 - Precision: FP16 - Batch Size: 32 - DLA Cores: Disabled Jetson AGX Xavier Jetson TX2 Max-P Jetson TX2 Max-Q 60 120 180 240 300 SE +/- 0.26, N = 3 SE +/- 0.07, N = 3 SE +/- 0.10, N = 3 259.82 41.91 32.67
NVIDIA TensorRT Inference Neural Network: AlexNet - Precision: INT8 - Batch Size: 4 - DLA Cores: Disabled OpenBenchmarking.org Images Per Second, More Is Better NVIDIA TensorRT Inference Neural Network: AlexNet - Precision: INT8 - Batch Size: 4 - DLA Cores: Disabled Jetson AGX Xavier Jetson TX2 Max-P Jetson TX2 Max-Q 200 400 600 800 1000 SE +/- 2.59, N = 3 SE +/- 2.79, N = 5 SE +/- 0.91, N = 3 1143 184 148
NVIDIA TensorRT Inference Neural Network: ResNet50 - Precision: FP16 - Batch Size: 4 - DLA Cores: Disabled OpenBenchmarking.org Images Per Second, More Is Better NVIDIA TensorRT Inference Neural Network: ResNet50 - Precision: FP16 - Batch Size: 4 - DLA Cores: Disabled Jetson AGX Xavier Jetson TX2 Max-P Jetson TX2 Max-Q 120 240 360 480 600 SE +/- 0.03, N = 3 SE +/- 1.32, N = 12 SE +/- 1.10, N = 12 547.50 92.28 72.01
TTSIOD 3D Renderer Phong Rendering With Soft-Shadow Mapping OpenBenchmarking.org FPS, More Is Better TTSIOD 3D Renderer 2.3b Phong Rendering With Soft-Shadow Mapping Jetson AGX Xavier Jetson TX2 Max-P Jetson TX2 Max-Q Raspberry Pi 3 Model B+ ASUS TinkerBoard Jetson TX1 Max-P 30 60 90 120 150 SE +/- 1.63, N = 12 SE +/- 0.15, N = 3 SE +/- 0.46, N = 4 SE +/- 0.16, N = 3 SE +/- 0.27, N = 9 SE +/- 0.04, N = 3 133.00 49.26 28.85 17.66 21.22 45.09 1. (CXX) g++ options: -O3 -fomit-frame-pointer -ffast-math -mtune=native -flto -lSDL -fopenmp -fwhole-program -lstdc++
NVIDIA TensorRT Inference Neural Network: ResNet50 - Precision: FP16 - Batch Size: 32 - DLA Cores: Disabled OpenBenchmarking.org Images Per Second, More Is Better NVIDIA TensorRT Inference Neural Network: ResNet50 - Precision: FP16 - Batch Size: 32 - DLA Cores: Disabled Jetson AGX Xavier Jetson TX2 Max-P Jetson TX2 Max-Q 140 280 420 560 700 SE +/- 1.23, N = 3 SE +/- 1.22, N = 3 SE +/- 0.86, N = 3 636.00 111.00 86.08
CUDA Mini-Nbody Test: Original OpenBenchmarking.org (NBody^2)/s, More Is Better CUDA Mini-Nbody 2015-11-10 Test: Original Jetson AGX Xavier Jetson TX2 Max-P Jetson TX2 Max-Q 11 22 33 44 55 SE +/- 0.00, N = 3 SE +/- 0.01, N = 3 SE +/- 0.03, N = 3 47.13 8.24 6.77
PyBench Total For Average Test Times OpenBenchmarking.org Milliseconds, Fewer Is Better PyBench 2018-02-16 Total For Average Test Times Jetson AGX Xavier Jetson TX2 Max-P Jetson TX2 Max-Q Raspberry Pi 3 Model B+ ASUS TinkerBoard Jetson TX1 Max-P 4K 8K 12K 16K 20K SE +/- 4.67, N = 3 SE +/- 33.86, N = 3 SE +/- 42.52, N = 3 SE +/- 43.80, N = 3 SE +/- 854.75, N = 9 SE +/- 18.55, N = 3 3007 5408 8735 20913 11502 6339
FLAC Audio Encoding WAV To FLAC OpenBenchmarking.org Seconds, Fewer Is Better FLAC Audio Encoding 1.3.2 WAV To FLAC Jetson AGX Xavier Jetson TX2 Max-P Jetson TX2 Max-Q Raspberry Pi 3 Model B+ ASUS TinkerBoard Jetson TX1 Max-P 70 140 210 280 350 SE +/- 0.61, N = 5 SE +/- 0.15, N = 5 SE +/- 0.18, N = 5 SE +/- 0.98, N = 5 SE +/- 2.51, N = 5 SE +/- 0.74, N = 5 54.47 65.07 104.28 339.53 279.05 79.20 1. (CXX) g++ options: -O2 -fvisibility=hidden -logg -lm
Zstd Compression Compressing ubuntu-16.04.3-server-i386.img, Compression Level 19 OpenBenchmarking.org Seconds, Fewer Is Better Zstd Compression 1.3.4 Compressing ubuntu-16.04.3-server-i386.img, Compression Level 19 Jetson AGX Xavier Jetson TX2 Max-P Jetson TX2 Max-Q Raspberry Pi 3 Model B+ ASUS TinkerBoard Jetson TX1 Max-P 110 220 330 440 550 SE +/- 0.91, N = 3 SE +/- 0.29, N = 3 SE +/- 1.02, N = 3 SE +/- 1.03, N = 3 SE +/- 2.16, N = 3 SE +/- 0.42, N = 3 80.06 144.97 253.80 342.23 496.62 145.80 1. (CC) gcc options: -O3 -pthread -lz -llzma
Rust Prime Benchmark Prime Number Test To 200,000,000 OpenBenchmarking.org Seconds, Fewer Is Better Rust Prime Benchmark Prime Number Test To 200,000,000 Jetson AGX Xavier Jetson TX2 Max-P Jetson TX2 Max-Q Raspberry Pi 3 Model B+ ASUS TinkerBoard Jetson TX1 Max-P 400 800 1200 1600 2000 SE +/- 0.00, N = 3 SE +/- 0.04, N = 3 SE +/- 0.09, N = 3 SE +/- 1.55, N = 3 SE +/- 187.90, N = 6 SE +/- 0.77, N = 3 32.37 104.96 170.25 1097.69 1821.05 128.45 1. (CC) gcc options: -pie -nodefaultlibs
C-Ray Total Time - 4K, 16 Rays Per Pixel OpenBenchmarking.org Seconds, Fewer Is Better C-Ray 1.1 Total Time - 4K, 16 Rays Per Pixel Jetson AGX Xavier Jetson TX2 Max-P Jetson TX2 Max-Q Raspberry Pi 3 Model B+ ASUS TinkerBoard Jetson TX1 Max-P 400 800 1200 1600 2000 SE +/- 7.17, N = 9 SE +/- 49.09, N = 9 SE +/- 1.44, N = 3 SE +/- 2.46, N = 3 SE +/- 22.09, N = 3 SE +/- 10.23, N = 3 355 585 869 2030 1718 753 1. (CC) gcc options: -lm -lpthread -O3
NVIDIA TensorRT Inference Neural Network: GoogleNet - Precision: FP16 - Batch Size: 32 - DLA Cores: Disabled OpenBenchmarking.org Images Per Second, More Is Better NVIDIA TensorRT Inference Neural Network: GoogleNet - Precision: FP16 - Batch Size: 32 - DLA Cores: Disabled Jetson AGX Xavier Jetson TX2 Max-P Jetson TX2 Max-Q 200 400 600 800 1000 SE +/- 0.21, N = 3 SE +/- 4.50, N = 3 SE +/- 2.17, N = 8 1006 233 179
NVIDIA TensorRT Inference Neural Network: AlexNet - Precision: FP16 - Batch Size: 32 - DLA Cores: Disabled OpenBenchmarking.org Images Per Second, More Is Better NVIDIA TensorRT Inference Neural Network: AlexNet - Precision: FP16 - Batch Size: 32 - DLA Cores: Disabled Jetson AGX Xavier Jetson TX2 Max-P Jetson TX2 Max-Q 400 800 1200 1600 2000 SE +/- 2.07, N = 3 SE +/- 7.68, N = 12 SE +/- 2.82, N = 3 2038 462 374
NVIDIA TensorRT Inference Neural Network: GoogleNet - Precision: FP16 - Batch Size: 4 - DLA Cores: Disabled OpenBenchmarking.org Images Per Second, More Is Better NVIDIA TensorRT Inference Neural Network: GoogleNet - Precision: FP16 - Batch Size: 4 - DLA Cores: Disabled Jetson AGX Xavier Jetson TX2 Max-P Jetson TX2 Max-Q 200 400 600 800 1000 SE +/- 2.48, N = 3 SE +/- 2.27, N = 3 SE +/- 1.90, N = 12 796 197 156
NVIDIA TensorRT Inference Neural Network: VGG19 - Precision: INT8 - Batch Size: 32 - DLA Cores: Disabled OpenBenchmarking.org Images Per Second, More Is Better NVIDIA TensorRT Inference Neural Network: VGG19 - Precision: INT8 - Batch Size: 32 - DLA Cores: Disabled Jetson AGX Xavier Jetson TX2 Max-P Jetson TX2 Max-Q 90 180 270 360 450 SE +/- 0.23, N = 3 SE +/- 0.06, N = 3 SE +/- 0.03, N = 3 394.66 15.92 12.59
NVIDIA TensorRT Inference Neural Network: VGG16 - Precision: INT8 - Batch Size: 32 - DLA Cores: Disabled OpenBenchmarking.org Images Per Second, More Is Better NVIDIA TensorRT Inference Neural Network: VGG16 - Precision: INT8 - Batch Size: 32 - DLA Cores: Disabled Jetson AGX Xavier Jetson TX2 Max-P Jetson TX2 Max-Q 100 200 300 400 500 SE +/- 0.10, N = 3 SE +/- 0.05, N = 3 SE +/- 0.01, N = 3 475.08 19.91 15.79
NVIDIA TensorRT Inference Neural Network: ResNet152 - Precision: INT8 - Batch Size: 32 - DLA Cores: Disabled OpenBenchmarking.org Images Per Second, More Is Better NVIDIA TensorRT Inference Neural Network: ResNet152 - Precision: INT8 - Batch Size: 32 - DLA Cores: Disabled Jetson AGX Xavier Jetson TX2 Max-P Jetson TX2 Max-Q 110 220 330 440 550 SE +/- 0.81, N = 3 SE +/- 0.03, N = 3 SE +/- 0.00, N = 3 493.22 22.07 17.36
NVIDIA TensorRT Inference Neural Network: ResNet50 - Precision: INT8 - Batch Size: 32 - DLA Cores: Disabled OpenBenchmarking.org Images Per Second, More Is Better NVIDIA TensorRT Inference Neural Network: ResNet50 - Precision: INT8 - Batch Size: 32 - DLA Cores: Disabled Jetson AGX Xavier Jetson TX2 Max-P Jetson TX2 Max-Q 300 600 900 1200 1500 SE +/- 0.25, N = 3 SE +/- 0.04, N = 3 SE +/- 0.08, N = 3 1215.08 59.69 47.15
NVIDIA TensorRT Inference Neural Network: ResNet152 - Precision: INT8 - Batch Size: 4 - DLA Cores: Disabled OpenBenchmarking.org Images Per Second, More Is Better NVIDIA TensorRT Inference Neural Network: ResNet152 - Precision: INT8 - Batch Size: 4 - DLA Cores: Disabled Jetson AGX Xavier Jetson TX2 Max-P Jetson TX2 Max-Q 80 160 240 320 400 SE +/- 1.59, N = 3 SE +/- 0.14, N = 3 SE +/- 0.15, N = 3 372.73 18.29 14.50
NVIDIA TensorRT Inference Neural Network: VGG19 - Precision: INT8 - Batch Size: 4 - DLA Cores: Disabled OpenBenchmarking.org Images Per Second, More Is Better NVIDIA TensorRT Inference Neural Network: VGG19 - Precision: INT8 - Batch Size: 4 - DLA Cores: Disabled Jetson AGX Xavier Jetson TX2 Max-P Jetson TX2 Max-Q 60 120 180 240 300 SE +/- 0.20, N = 3 SE +/- 0.25, N = 4 SE +/- 0.23, N = 3 265.81 14.32 11.45
NVIDIA TensorRT Inference Neural Network: ResNet50 - Precision: INT8 - Batch Size: 4 - DLA Cores: Disabled OpenBenchmarking.org Images Per Second, More Is Better NVIDIA TensorRT Inference Neural Network: ResNet50 - Precision: INT8 - Batch Size: 4 - DLA Cores: Disabled Jetson AGX Xavier Jetson TX2 Max-P Jetson TX2 Max-Q 200 400 600 800 1000 SE +/- 1.86, N = 3 SE +/- 0.79, N = 4 SE +/- 0.64, N = 3 902.78 49.97 39.15
NVIDIA TensorRT Inference Neural Network: VGG16 - Precision: INT8 - Batch Size: 4 - DLA Cores: Disabled OpenBenchmarking.org Images Per Second, More Is Better NVIDIA TensorRT Inference Neural Network: VGG16 - Precision: INT8 - Batch Size: 4 - DLA Cores: Disabled Jetson AGX Xavier Jetson TX2 Max-P Jetson TX2 Max-Q 70 140 210 280 350 SE +/- 0.46, N = 3 SE +/- 0.25, N = 6 SE +/- 0.20, N = 5 303.78 17.56 14.24
OpenCV Benchmark OpenBenchmarking.org Seconds, Fewer Is Better OpenCV Benchmark 3.3.0 Jetson AGX Xavier Jetson TX2 Max-P Jetson TX2 Max-Q Raspberry Pi 3 Model B+ 110 220 330 440 550 SE +/- 1.57, N = 3 SE +/- 0.27, N = 3 SE +/- 5.74, N = 3 128.00 296.00 493.00 2.74 1. (CXX) g++ options: -std=c++11 -rdynamic
NVIDIA TensorRT Inference Neural Network: GoogleNet - Precision: INT8 - Batch Size: 32 - DLA Cores: Disabled OpenBenchmarking.org Images Per Second, More Is Better NVIDIA TensorRT Inference Neural Network: GoogleNet - Precision: INT8 - Batch Size: 32 - DLA Cores: Disabled Jetson AGX Xavier Jetson TX2 Max-P Jetson TX2 Max-Q 400 800 1200 1600 2000 SE +/- 8.72, N = 3 SE +/- 0.74, N = 3 SE +/- 0.07, N = 3 1693 130 104
NVIDIA TensorRT Inference Neural Network: AlexNet - Precision: INT8 - Batch Size: 32 - DLA Cores: Disabled OpenBenchmarking.org Images Per Second, More Is Better NVIDIA TensorRT Inference Neural Network: AlexNet - Precision: INT8 - Batch Size: 32 - DLA Cores: Disabled Jetson AGX Xavier Jetson TX2 Max-P Jetson TX2 Max-Q 700 1400 2100 2800 3500 SE +/- 1.06, N = 3 SE +/- 0.52, N = 3 SE +/- 1.39, N = 3 3143 301 237
NVIDIA TensorRT Inference Neural Network: GoogleNet - Precision: INT8 - Batch Size: 4 - DLA Cores: Disabled OpenBenchmarking.org Images Per Second, More Is Better NVIDIA TensorRT Inference Neural Network: GoogleNet - Precision: INT8 - Batch Size: 4 - DLA Cores: Disabled Jetson AGX Xavier Jetson TX2 Max-P Jetson TX2 Max-Q 200 400 600 800 1000 SE +/- 4.31, N = 3 SE +/- 1.65, N = 3 SE +/- 1.32, N = 3 1146.00 113.00 88.88
Tesseract OCR Time To OCR 7 Images OpenBenchmarking.org Seconds, Fewer Is Better Tesseract OCR 4.0.0-beta.1 Time To OCR 7 Images Jetson AGX Xavier 16 32 48 64 80 SE +/- 0.89, N = 3 71.94
LeelaChessZero Backend: CUDA + cuDNN FP16 OpenBenchmarking.org Nodes Per Second, More Is Better LeelaChessZero 0.20.1 Backend: CUDA + cuDNN FP16 Jetson AGX Xavier 500 1000 1500 2000 2500 SE +/- 7.60, N = 3 2515.01 1. (CXX) g++ options: -lpthread -lz
LeelaChessZero Backend: CUDA + cuDNN OpenBenchmarking.org Nodes Per Second, More Is Better LeelaChessZero 0.20.1 Backend: CUDA + cuDNN Jetson AGX Xavier 200 400 600 800 1000 SE +/- 6.14, N = 3 952.89 1. (CXX) g++ options: -lpthread -lz
LeelaChessZero Backend: BLAS OpenBenchmarking.org Nodes Per Second, More Is Better LeelaChessZero 0.20.1 Backend: BLAS Jetson AGX Xavier 11 22 33 44 55 SE +/- 0.62, N = 7 47.62 1. (CXX) g++ options: -lpthread -lz
GLmark2 Resolution: 1920 x 1080 OpenBenchmarking.org Score, More Is Better GLmark2 Resolution: 1920 x 1080 Jetson AGX Xavier 600 1200 1800 2400 3000 2876
NVIDIA TensorRT Inference Neural Network: AlexNet - Precision: FP16 - Batch Size: 4 - DLA Cores: Disabled OpenBenchmarking.org Images Per Second, More Is Better NVIDIA TensorRT Inference Neural Network: AlexNet - Precision: FP16 - Batch Size: 4 - DLA Cores: Disabled Jetson AGX Xavier Jetson TX2 Max-P Jetson TX2 Max-Q 300 600 900 1200 1500 SE +/- 1.82, N = 3 SE +/- 7.77, N = 12 SE +/- 3.03, N = 6 1200 264 216
Phoronix Test Suite v10.8.4