Blog Post

Azure High Performance Computing (HPC) Blog
2 MIN READ

Announcing Azure HBv5 Virtual Machines: A Breakthrough in Memory Bandwidth for HPC

Fernando_Aznar's avatar
Nov 19, 2024

The next generation of purpose-built HPC Azure virtual machines

Today during Microsoft Ignite, Satya Nadella unveiled our latest CPU-based virtual machine for HPC customers and their applications, Azure HBv5. This new virtual machine (VM) is optimized for the most memory bandwidth-intensive HPC applications, such as computational fluid dynamics, automotive and aerospace simulation, weather modeling, energy research, molecular dynamics, computer aided engineering, and more.

8x higher memory bandwidth to solve the biggest HPC bottleneck

For many HPC customers, memory performance from standard server designs has become the most significant impediment to achieving desired levels of workload performance (time to insight) and cost efficiency. To overcome this bottleneck, Microsoft and AMD have worked together to develop a custom 4th Generation EPYC™ processor with high bandwidth memory (HBM). In an Azure HBv5 VM, four of these processors work jointly to deliver nearly 7 TB/s of memory bandwidth. For comparison, this is up to 8x higher compared to the latest bare-metal and Cloud alternatives, almost 20x more than Azure HBv3 and Azure HBv2 (3rd Gen EPYC™ with 3D V-cache “Milan-X,” and 2nd Gen EPYC™ “Rome”), and up to 35x more than a 4–5-year-old HPC server approaching the end of its hardware lifecycle.

Below chart shows STREAM performance for an Azure HBv5 VM compared to prior generations of Azure H-series VMs.

Figure 1: STREAM Triad memory bandwidth for Azure HBv5 and all current Azure H-series VMs available in Azure

HPC Innovations and optimizations across the technology stack

While Azure HBv5’s memory bandwidth is a standout feature, Microsoft and AMD have co-engineered improvements throughout the product to deliver customers a VM that is balanced, secure, configurable to user needs, and extremely performant for a variety of HPC workloads.

Each Azure HBv5 VM will feature:

  • 6.9 TB/s of memory bandwidth (STREAM Triad) across 400-450 GB of RAM (HBM3)
  • Up to 9 GB of memory per core (customer configurable)
  • Up to 352 AMD EPYC “Zen4” CPU cores, 4 GHz peak frequencies (customer configurable)
  • 2X total Infinity Fabric bandwidth among CPUs as any AMD EPYC™ server platform to date
  • SMT disabled, single-tenant only design (1 VM per server)
  • 800 Gb/s of NVIDIA Quantum-2 InfiniBand, balanced as 200 Gb/s per CPU SoC
  • Azure VMSS Flex to scale MPI workloads to hundreds of thousands of HBM-powered CPU cores
  • 160 Gbps of Azure Accelerated Networking via 2nd generation Azure Boost NIC
  • 14 TB local NVMe SSD delivering up to 50 GB/s read and 30 GB/s write bandwidth

Sign up for the Preview of Azure HBv5 virtual machines

Starting today, you can sign up for the Azure HBv5 Preview which will start in the first half of 2025. Stop by Microsoft Azure booth #1905 at Supercomputing 2024 in Atlanta, Georgia November 19-22 to see Azure HBv5 alongside other Azure supercomputing products, and chat with experts about what this VM can do for your HPC workloads.

Figure 2: A Microsoft Azure HBv5 server, viewable at the Supercomputing 2024 Conference

Published Nov 19, 2024
Version 1.0
No CommentsBe the first to comment