'HB' sub-family storage optimized VM size series

Applies to: ✔️ Linux VMs ✔️ Windows VMs ✔️ Flexible scale sets ✔️ Uniform scale sets

The 'HB' sub-family of VM size series are one of Azure's high-performance computing (HPC) optimized H-family VM instances. They're designed for compute-intensive workloads, such as computational fluid dynamics, finite element analysis, and large-scale scientific simulations. HB-series VMs, equipped with high-performance AMD EPYC processors and fast memory, deliver exceptional CPU and memory bandwidth. They're ideal for applications that need extensive computational resources to handle large-scale calculations and data processing. They're well-suited for industries such as engineering, scientific research, and data analysis. In these fields, processing speed and accuracy are essential for maintaining productivity and driving innovation.

Workloads and use cases

Computational Fluid Dynamics (CFD): HB family VMs are ideal for simulations in fields like aerospace, automotive design, and manufacturing, where fluid dynamics calculations are intensive.

Finite Element Analysis (FEA): HB family VMs are suitable for engineering analyses that simulate physical phenomena, requiring intensive computational power to model complex systems and materials.

Weather Forecasting: HB family VMs can handle the massive datasets and complex simulations required for high-resolution weather modeling and forecasting.

Seismic Processing: HB family VMs are in the oil and gas industry, HB family VMs can process seismic data to help map and understand subsurface structures.

Scientific Research: HB family VMs support a wide range of scientific research that requires large-scale mathematical modeling, including physics and computational chemistry simulations.

Genomics and Bioinformatics: HB family VMs are also used in life sciences for genomic analysis, where large amounts of data need to be processed quickly to decode genetic information.

Series in family

HBv2-series

HBv2-series VMs are optimized for applications that are driven by memory bandwidth, such as fluid dynamics, finite element analysis, and reservoir simulation. HBv2 VMs feature 120 AMD EPYC 7V12 processor cores, 4 GB of RAM per CPU core, and no simultaneous multithreading. Each HBv2 VM provides up to 350 GB/s of memory bandwidth, and up to 4 teraFLOPS of FP64 compute. HBv2-series VMs feature 200 Gb/sec Mellanox HDR InfiniBand. These VMs are connected in a non-blocking fat tree for optimized and consistent RDMA performance. These VMs support Adaptive Routing and the Dynamic Connected Transport (DCT, in addition to standard RC and UD transports). These features enhance application performance, scalability, and consistency, and their usage is recommended.

View the full hbv2-series page.

HBv3-series

HBv3-series Virtual Machines (VMs) are designed for very demanding computing tasks. They help with things like studying how liquids move, analyzing structures, predicting weather, and processing earthquake data. They also support oil reservoir modeling and testing computer chip designs. HBv3 VMs feature up to 120 AMD EPYC™ 7V73X (Milan-X) CPU cores, 448 GB of RAM, and no simultaneous multithreading. HBv3-series VMs also provide 350 GB/sec of memory bandwidth (amplified up to 630 GB/s), up to 96 MB of L3 cache per core (1536 MB total per VM), up to 7 GB/s of block device SSD performance, and clock frequencies up to 3.5 GHz. All HBv3-series VMs feature 200 Gb/sec High Data Rate (HDR) InfiniBand from NVIDIA Networking to enable supercomputer-scale Message Pass Interface (MPI) workloads. These VMs are connected in a nonblocking fat tree for optimized and consistent RDMA performance. The HDR InfiniBand fabric also supports Adaptive Routing and Dynamic Connected Transport (DCT), along with standard Reliable Connection (RC) and Unreliable Datagram (UD) transports. These features improve performance, make scaling easier, and keep results consistent.

View the full hbv3-series page.

HBv4-series

HBv4-series VMs are optimized for many high-performance computing (HPC) workloads. These include:

• Computational fluid dynamics
• Finite element analysis
• Frontend and backend Electronic Design Automation (EDA)
• Rendering and graphics work
• Molecular dynamics
• Computational geoscience
• Weather simulation
• Financial risk analysis

HBv4 VMs feature up to 176 AMD EPYC™ 9V33X ("Genoa-X") CPU cores with AMD's 3D V-Cache, clock frequencies up to 3.7 GHz, and no simultaneous multithreading. HBv4-series VMs also provide 768 GB of RAM, 2304 MB L3 cache. Each VM has 2304 MB of L3 cache that delivers up to 5.7 TB/s of bandwidth. This boosts DRAM bandwidth of 780 GB/s, giving an average of 1.2 TB/s of effective memory speed for many workloads. The VMs also provide up to 12 GB/s (reads) and 7 GB/s (writes) of block device SSD performance.

All HBv4-series VMs feature 400 Gb/s NDR InfiniBand from NVIDIA Networking to enable supercomputer-scale MPI workloads. These VMs are connected in a non-blocking fat tree for optimized and consistent RDMA performance. NDR continues to support features like Adaptive Routing and the Dynamically Connected Transport (DCT). This newest generation of InfiniBand also brings greater support for offload of MPI collectives, optimized real-world latencies due to congestion control intelligence, and enhanced adaptive routing capabilities. These features enhance application performance, scalability, and consistency, and their usage is recommended.

View the full hbv4-series page.

HBv5-series

HBv5-series VMs are optimized for the most memory bandwidth-intensive HPC applications, including:

• Computational fluid dynamics
• Automotive and aerospace simulations
• Weather modeling
• Energy research
• Molecular dynamics simulations
• Computer-aided engineering and other HPC workloads

HBv5 VMs feature 6.7 TB/s of memory bandwidth across 432 GB of high-bandwidth memory (HBM) and up to 368 4th Generation AMD EPYC™ processor cores with 4 GHz boost frequencies, 3.5 GHz base frequencies, and no simultaneous multithreading.

Each HBv5-series VM also includes 14.3 TiB of local NVMe SSD storage with up to 50 GB/s (reads) and 30 GB/s (writes) of block device performance.

All HBv5-series VMs feature 800 Gb/s per node (4 x 200 Gb/s CX-7 NIC) of InfiniBand connectivity from NVIDIA Networking to enable supercomputer-scale MPI workloads. These VMs are connected in a nonblocking fat tree for optimized and consistent RDMA performance.

The InfiniBand NICs support features like Adaptive Routing, Dynamically Connected Transport (DCT), hardware acceleration of MPI collectives, and congestion control. These features enhance application performance, scalability, and consistency, and their usage is recommended for optimal performance and cost efficiency.

View the full hbv5-series page.

Other size information

List of all available sizes: Sizes

Pricing Calculator: Pricing Calculator

Information on Disk Types: Disk Types

Next steps

Take advantage of the latest performance and features available for your workloads by changing the size of a virtual machine.

Utilize Microsoft's in-house designed ARM processors with Azure Cobalt VMs.

Learn how to Monitor Azure virtual machines.