Ubuntu 16.04.6 LTS TITAN RTX x2 RTX2080Ti x2 CUDA 10.2 E5-2650 V4 x2 NVIDIA GPU CLOUD 18.04-py3 tensorflow x4GPU googlenet x2GPU googlenet 学習能力を比較してみた Images/sec 3879.4 1884.2

参照サイト

TensorFlow: 1.7.0
This script: nvcnn.py v1.6
Cmd line args:
–model=googlenet
–batch_size=256
–num_gpus=4
–fp16
Num images: Synthetic
Model: googlenet
Batch size: 1024 global
256.0 per device
Devices: [‘/gpu:0’, ‘/gpu:1’, ‘/gpu:2’, ‘/gpu:3’]　TITANRTX x2 TITANRTX x2 4GPU

スコア　Images/sec 3879.4

TensorFlow: 1.7.0
This script: nvcnn.py v1.6
Cmd line args:
–model=googlenet
–batch_size=256
–num_gpus=2
–fp16
Num images: Synthetic
Model: googlenet
Batch size: 512 global
256.0 per device
Devices: [‘/gpu:0’, ‘/gpu:1’] TITANRTX x2 2GPU

スコア　Images/sec 1884.2

データ詳細 Ubuntu 16.04.6 LTS TITAN RTX x2 RTX2080Ti x2 CUDA 10.2 E5-2650 V4 x2 NVIDIA GPU CLOUD 18.04-py3 tensorflow x4GPU googlenet x2GPU googlenet 学習能力確認 Images sec 3879.4 1884.2

GPU温度推移 Ubuntu 16.04.6 LTS TITAN RTX x2 RTX2080Ti x2 CUDA 10.2 E5-2650 V4 x2 NVIDIA GPU CLOUD 18.04-py3 tensorflow x4GPU googlenet x2GPU googlenet 学習能力確認 Images sec 3879.4 1884.2 nvidia-smi