Providing GPU-accelerated SW configuration containers
Deep Learning Containers, Updated Monthly
Azure NCv2, NCv3, and ND support
On August 30, NVIDIA announced that it will support NVIDIA GPU Cloud (NGC) on Microsoft's cloud platform 'Azure'.

With ready-to-run containers from NGC, supported on Azure, developers can eliminate the complexity of software integration and testing. They can access on-demand GPU computing that scales as needed.


Faster AI and HPC Projects Now Possible
Building and testing software stacks to run widely used deep learning software such as TensorFlow, Microsoft Cognitive Toolkit, PyTorch, and NVIDIA TensorRT is challenging and time-consuming.

Additionally, they have dependencies on the operating system (OS) level, drivers, libraries, and runtimes, and many packages recommend different versions of supported components. Additionally, frameworks and applications are updated frequently, so this task must be repeated every time a new version is released.

Ideally, you should test the new version to ensure it performs as well or better than the previous version, and this process should be done before starting the project.

For HPC, deploying the latest software across a cluster of systems is a challenging task. This involves finding, installing, and testing the correct dependencies, and this task must be performed across multiple systems in a multi-tenant environment.

NGC eliminates this complexity by providing pre-configured containers with GPU-accelerated software. As a result of NVIDIA's ongoing R&D investment, deep learning containers enable containers to leverage the latest GPU capabilities. Furthermore, NVIDIA tests, tunes, and optimizes the entire software stack within these deep learning containers, which are updated monthly to ensure optimal performance.

NVIDIA contributes to open source projects by working closely with communities and framework developers. In 2017, NVIDIA contributed to 800 open source projects. We are also collaborating with other container developers on NGC to optimize applications and test performance and compatibility.


Running Azure Instances with NVIDIA GPUs
Users can run the following Microsoft Azure instances with NVIDIA GPUs, accessing 35 GPU-accelerated containers for deep learning software, HPC applications, HPC visualization tools, and a variety of partner applications from the NGC container registry:

▲NCv3 (1, 2, or 4 NVIDIA Tesla V100 GPUs), ▲NCv2 (1, 2, or 4 NVIDIA Tesla P100 GPUs), ▲ND (1, 2, or 4 NVIDIA Tesla P40 GPUs)

The same NGC container runs on Azure instances even if they have different types or quantities of GPUs.


How to use NGC containers in Azure
Click on the NVIDIA GPU Cloud Image for Deep Learning and HPC in the Microsoft Azure Marketplace to launch a compatible NVIDIA GPU instance on Azure. Then, import the desired containers from the NGC registry onto the running instance. Before proceeding, you must sign up for a free NGC account. For more information, see the "Using NGC Containers on Azure" documentation.

In addition to clicking on the NVIDIA images in the Azure Marketplace to run these NGC containers, you can also download and run Azure Batch AI from NGC on VMs such as Azure NCv2, NCv3, and ND. You can get started with Azure Batch AI with NGC containers by following the GitHub instructions.

Meanwhile, NVIDIA will host a webinar related to this announcement at 1:00 AM Korean time on October 2nd.