output
2 x Intel Xeon E5-2620 v4 testing with a Supermicro X10DRG-O(T)+-CPU v1.00 (3.1 BIOS) and llvmpipe 11GB on Arch rolling via the Phoronix Test Suite.


llvmpipe: 

	Processor: 2 x Intel Xeon E5-2620 v4 @ 2.10GHz (16 Cores / 32 Threads), Motherboard: Supermicro X10DRG-O(T)+-CPU v1.00 (3.1 BIOS), Chipset: Intel Xeon E7 v4/Xeon, Memory: 64512MB, Disk: 2 x 1000GB Samsung SSD 860 + 1000GB Seagate ST1000NX0313, Graphics: llvmpipe 11GB, Audio: NVIDIA TU102 HD Audio, Monitor: ASUS VH242H, Network: 2 x Intel 10-Gigabit X540-AT2

	OS: Arch rolling, Kernel: 5.3.6-arch1-1-ARCH (x86_64), Desktop: KDE Plasma 5.16.5, Display Server: X Server 1.20.5, Display Driver: modesetting 1.20.5, OpenGL: 3.3 Mesa 19.2.1 (LLVM 9.0 256 bits), Compiler: GCC 9.2.0 + CUDA 10.1, File-System: xfs, Screen Resolution: 1920x1080


NVIDIA GPU Cloud TensorFlow 18.09
Test: VGG-16, FP16
Images Per Second > Higher Is Better
llvmpipe . 205.60 |============================================================


NVIDIA GPU Cloud TensorFlow 18.09
Test: VGG-16, FP32
Images Per Second > Higher Is Better
llvmpipe . 149.00 |============================================================


NVIDIA GPU Cloud TensorFlow 18.09
Test: AlexNet, FP16
Images Per Second > Higher Is Better
llvmpipe . 4278.77 |===========================================================


NVIDIA GPU Cloud TensorFlow 18.09
Test: AlexNet, FP32
Images Per Second > Higher Is Better
llvmpipe . 3166.73 |===========================================================


NVIDIA GPU Cloud TensorFlow 18.09
Test: Googlenet, FP16
Images Per Second > Higher Is Better
llvmpipe . 937.47 |============================================================


NVIDIA GPU Cloud TensorFlow 18.09
Test: Googlenet, FP32
Images Per Second > Higher Is Better
llvmpipe . 725.80 |============================================================


NVIDIA GPU Cloud TensorFlow 18.09
Test: ResNet-50, FP16
Images Per Second > Higher Is Better
llvmpipe . 406.03 |============================================================


NVIDIA GPU Cloud TensorFlow 18.09
Test: ResNet-50, FP32
Images Per Second > Higher Is Better
llvmpipe . 261.57 |============================================================


NVIDIA GPU Cloud TensorFlow 18.09
Test: Inception v4, FP16
Images Per Second > Higher Is Better
llvmpipe . 106.77 |============================================================


NVIDIA GPU Cloud TensorFlow 18.09
Test: Inception v4, FP32
Images Per Second > Higher Is Better
llvmpipe . 76.83 |=============================================================