New Azure NC H100 v5 VMs Optimized for Generative AI and HPC workloads is now Generally Available
Published Mar 18 2024 03:00 PM 2,521 Views

Azure NC H100 v5 virtual machines (VMs) are an excellent platform for executing diverse AI and High-Performance Computing (HPC) workloads. These workloads demand substantial computational power, large capacity of high-performance memory, and advanced GPU acceleration. In addition to AI, the Azure NC H100 v5 VMs are particularly well-suited for extreme modelling and simulation demands in the following science and mathematics disciplines: Computational Fluid Dynamics (CFD), Molecular Dynamics, Quantum Chemistry, Weather Forecasting and Climate Modeling, and Financial Analytics.


The AI landscape is constantly expanding and evolving, moving at a dizzying pace. Generative AI technology has played a pivotal role, enabling a diverse array of use cases. These range from powering AI assistants, chatbots, and search engines to facilitating creative content generation. As Generative AI applications expand at incredible speed, the fundamental language models that empower them will expand also to include both Small Language Models (SMLs) and Large Language Models (LLMs). In addition, Artificial Narrow Intelligence (ANI) models will continue to evolve focused on more precise predictions rather than creation of novel data to continue to enhance its use cases. Their applications include tasks such as image classification, object detection, and broader natural language processing.


At Microsoft, our mission is to empower every person and every organization on the planet to achieve more.  Leveraging the robust capabilities and scalability of Microsoft Azure, we offer computational tools that empower organizations of all sizes, regardless of their resources. Azure NC H100 v5 VMs is yet another computational tool made generally available today that will do just that.


Here are some examples of what our customers are doing with our existing NC-series VMs and planning with the power of Azure NC H100 v5 GPU Virtual Machines at their fingertips:


Snorkel AI is a Microsoft for Startups Pegasus partner that helps enterprises move AI projects from prototype to production. A founding member of the Stanford Center for Research on Foundation Models, Snorkel AI is grounded in years of academic research and endeavors to remain at the forefront of new scholarship in data-centric AI and foundation models.

"Snorkel’s recent top tier ranking on the AlpacaEval 2.0 LLM leaderboard would not have been possible without the Microsoft for Startups Pegasus Program. Access to SoTA NVIDIA A100s via a seamless Azure experience has empowered us to drive cutting-edge research in programmatic alignment/DPO in a quick & efficient manner. For example, Azure AI Infrastructure VMs allow our research team to run quick experiments from small projects to large-scale distributed jobs reliably and with full monitoring mechanisms. Designing research projects on Azure’s next generation NC H100 v5-series powered by NVIDIA H100 NVL PCIe GPU will help our researchers deliver value for our customers and the OSS community even faster.”- Hoang Tran, Senior Research Scientist, Snorkel AI


Northflank is a self-service developer platform that automates and unifies deployment of any workload, on any cloud, at any scale.

"Northflank's customers want to build scale-out apps on top of a sizeable number of Azure NC H100 v5 series VMs that feature the NVIDIA 100 GPUs running in a Kubernetes clusters, while keeping the self-service experience that Northflank’s developer platform provides. With NC H100 v5, Azure is the fastest way for us to help those customers ship apps on scale-out GPU infrastructure." - Will Stewart, CEO & Co-Founder, Northflank


SLB is a global technology company driving energy innovation for a balanced planet.

“The Azure NC H100 v5 VM series is now able to deliver a step change in performance for Delfi Seismic Processing and Imaging workloads by leveraging the NVIDIA H100 GPUs. Following SLB feedback, Microsoft incorporated our high-performance compute requirements for seismic imaging into the hardware design.” - Robert Bloor, Seismic Processing & Imaging Science and Innovation Manager, SLB


Introducing the new NC H100 v5 series virtual machine, now generally available

Today, we are excited to announce that Azure NC H100 v5 Virtual Machines are now generally available. The NC H100 v5-series virtual machine (VM) is a cutting-edge addition to the Azure GPU virtual machines family. Designed for mid-range AI model training and generative inferencing, and HPC simulation workloads. This series combines the power of NVIDIA H100 NVL GPUs with 4th-generation AMD EPYC™ Genoa processors.


The NC H100 v5-series offers two classes of virtual machines, ranging from one to two NVIDIA H100 94GB NVL Tensor Core GPUs. It is more cost-effective than ever before, while still giving customers the options and flexibility they need for their workloads. We can’t wait to see what you’ll build, analyze, and discover with the new Azure NC H100 v5 platform.


For AI Inference workloads, customers will experience between 1.6x-1.9x inference performance on one GPU size depending on the types of workloads. The NC H100 v5 VMs offer significant performance improvements over the previous generations of Azure VMs in the NC series. The H100 NVL PCIe GPUs provide up to 2x the compute performance, 2x the memory bandwidth, and 17% larger HBM GPU memory capacity per VM compared to the A100 GPUs. The H100 NVL PCIe GPUs support PCIe Gen5, which provides the highest communication speeds (128GB/s bi-directional) between the host processor and the GPU. This reduces the latency and overhead of data transfer and enables faster and more scalable AI and HPC applications.


The NC H100 v5-series VMs empower your AI and HPC workloads, providing the performance and flexibility you need. Whether you’re training models, running inferencing tasks, or developing cutting-edge applications, these VMs have you covered. Explore the future of AI with the NC H100 v5-series on Azure!


Learn more

Version history
Last update:
‎Mar 26 2024 10:11 AM
Updated by: