Jetson TX1 vs. TX2 Caffe

Tests for a future article.

HTML result view exported from: https://openbenchmarking.org/result/1703144-RI-TEGRACAFF40&gru.

Jetson TX1 vs. TX2 CaffeProcessorMotherboardMemoryDiskGraphicsOSKernelDesktopDisplay ServerDisplay DriverOpenGLVulkanCompilerFile-SystemScreen ResolutionJetson TX1Jetson TX2 Max-QJetson TX2 Max-PARMv8 rev 1 @ 1.73GHz (4 Cores)jetson_tx14096MB16GB 016G32NVIDIA Tegra X1 (nvgpu)/Ubuntu 16.043.10.96-tegra (aarch64)Unity 7.4.0X Server 1.18.3NVIDIA 24.2.14.5.01.0.8GCC 5.4.0 20160609 + CUDA 8.0ext43840x2160ARMv8 rev 3 @ 2.00GHz (6 Cores)quill8192MB31GB 032G34NVIDIA TEGRA4.4.15-tegra (aarch64)NVIDIA 1.0.0OpenBenchmarking.orgEnvironment Details- Jetson TX1: __GL_PERFMON_MODE=1Compiler Details- --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 TX1: Scaling Governor: tegra interactive- Jetson TX2 Max-Q: Scaling Governor: tegra_cpufreq schedutil- Jetson TX2 Max-P: Scaling Governor: tegra_cpufreq schedutil

Jetson TX1 vs. TX2 Caffecaffe: CUDA AlexNetcaffe: CUDA GooglenetJetson TX1Jetson TX2 Max-QJetson TX2 Max-P204205431604179031382371143144301567OpenBenchmarking.org

Caffe

Build: CUDA AlexNet

OpenBenchmarking.orgMilli-Seconds, Fewer Is BetterCaffe 2016-12-29Build: CUDA AlexNetJetson TX1Jetson TX2 Max-QJetson TX2 Max-P40K80K120K160K200KSE +/- 1130.47, N = 3SE +/- 75.77, N = 3SE +/- 117.75, N = 32042051790311431441. (CXX) g++ options: -pthread -fPIC -O2 -lcaffe -lglog -lgflags -lprotobuf -lboost_system -lboost_filesystem -lm -lhdf5_hl -lhdf5 -lleveldb -lsnappy -llmdb -lopencv_core -lopencv_highgui -lopencv_imgproc -lboost_thread -lstdc++ -lcblas -latlas

Caffe

Build: CUDA Googlenet

OpenBenchmarking.orgMilli-Seconds, Fewer Is BetterCaffe 2016-12-29Build: CUDA GooglenetJetson TX1Jetson TX2 Max-QJetson TX2 Max-P90K180K270K360K450KSE +/- 369.46, N = 3SE +/- 395.20, N = 3SE +/- 266.95, N = 34316043823713015671. (CXX) g++ options: -pthread -fPIC -O2 -lcaffe -lglog -lgflags -lprotobuf -lboost_system -lboost_filesystem -lm -lhdf5_hl -lhdf5 -lleveldb -lsnappy -llmdb -lopencv_core -lopencv_highgui -lopencv_imgproc -lboost_thread -lstdc++ -lcblas -latlas

Phoronix Test Suite v10.8.4