Tensorflow Tesla® instances benchmark

Summary of test model results for the images classification with Tesla® LeaderGPU® servers

LeaderGPU® is a new player in the GPU computing market, and it intends to change the rules of the game. At this moment in time, the GPU computing market comprises several large players such as Amazon AWS, Google Cloud, etc. However, a large player does not always mean the best market offer. The LeaderGPU® project, in comparison to Amazon AWS and Google Cloud, provides physical servers, not VPS, where hardware resources can be shared among several dozens of users.

Tests were conducted on the LeaderGPU® Tesla® computing systems on synthetic data of the following network models: ResNet-50, ResNet-152, VGG16 and AlexNet. At the end of this article, you will find the results of tests carried out on other models. The testing of synthetic data was performed using tf.Variable in analogy with the models configured for ImageNet.

The following commands were used to run the test:

git clone https://github.com/tensorflow/benchmarks.git

python3.5 benchmarks/scripts/tf_cnn_benchmarks/tf_cnn_benchmarks.py --num_gpus=2 --model alexnet (vgg11, vgg16, etc.) --batch_size 32 (64, 128, 256, 512)

LeaderGPU® Tesla® instances

Testing environment: 2 x Tesla® P100 PCI (ltbv32), 2 x Tesla® V100 PCI (ltbv20), 2 x Tesla® V100 NVLink™ (ltbv46)

Instance type: 2 x Tesla® P100 PCI (ltbv32), 2 x Tesla® V100 PCI (ltbv20), 2 x Tesla® V100 NVLink (ltbv46) 

GPUs: Nvidia® Tesla® cards OS: CentOS 7 CUDA® / cuDNN: 9.0 / 7.0.5 TensorFlow™ 1.7 from repo Benchmark GitHub hash: 9165a70 Date of testing: 25.04.2018

Other results

Testing synthetic data (images / s)

2x PCI Tesla® P100