Caffe vs. nvCaffe On Jetson TX2
ARMv8 rev 3 testing with a quill and GP10B (nvgpu)/ on Ubuntu 16.04 via the Phoronix Test Suite. nvCaffe modifications. Tests for a future article


Caffe: 

	Processor: ARMv8 rev 3 @ 2.00GHz (6 Cores), Motherboard: quill, Memory: 8192MB, Disk: 31GB 032G34, Graphics: GP10B (nvgpu)/

	OS: Ubuntu 16.04, Kernel: 4.4.15-tegra (aarch64), Desktop: Unity 7.4.0, Display Server: X Server 1.18.3, Display Driver: NVIDIA 27.1.0, OpenGL: 4.5.0, Vulkan: 1.0.8, Compiler: GCC 5.4.0 20160609 + CUDA 8.0, File-System: ext4, Screen Resolution: 3840x2160

NVIDIA nvCaffe: 

	Processor: ARMv8 rev 3 @ 2.00GHz (6 Cores), Motherboard: quill, Memory: 8192MB, Disk: 31GB 032G34, Graphics: GP10B (nvgpu)/

	OS: Ubuntu 16.04, Kernel: 4.4.15-tegra (aarch64), Desktop: Unity 7.4.0, Display Server: X Server 1.18.3, Display Driver: NVIDIA 27.1.0, OpenGL: 4.5.0, Vulkan: 1.0.8, Compiler: GCC 5.4.0 20160609 + CUDA 8.0, File-System: ext4, Screen Resolution: 3840x2160


Caffe 2016-12-29
Build: CUDA AlexNet
Milli-Seconds < Lower Is Better
Caffe .......... 143994 |======================================================
NVIDIA nvCaffe . 120787 |=============================================


Caffe 2016-12-29
Build: CUDA Googlenet
Milli-Seconds < Lower Is Better
Caffe .......... 301424 |======================================================
NVIDIA nvCaffe . 237643 |===========================================