We are excited to announce Azure HBv4 and HX-series Virtual Machines (VMs) at the moment are usually obtainable. With the overall availability, Microsoft is providing prospects the primary VMs that includes the most recent 4th Gen AMD EPYC™ processors with AMD 3D V-Cache™ know-how (codename ‘Genoa-X’), paired with 400 Gigabit NVIDIA Quantum-2 InfiniBand. Azure HBv4 and HX-series VMs supply management ranges of efficiency, scaling effectivity, and cost-effectiveness for a wide range of HPC workloads corresponding to computational fluid dynamics (CFD), monetary providers calculations, finite component evaluation (FEA), geoscience simulations, climate simulation, rendering, quantum chemistry, and silicon design.
Compared to Azure HBv3-series VMs utilizing 3rd Gen AMD EPYC™ processors (codename ‘Milan-X’), already the highest efficiency VMs for HPC workloads on the general public cloud, prospects will see as much as:
- 1.6 instances greater efficiency for rendering
- 2.4 instances greater efficiency for climate simulation
- 2.7 instances greater efficiency for CFD
- 4.2 instances greater efficiency for molecular dynamics
- 5.7 instances greater efficiency for structural evaluation
Visit our technical weblog for extra detailed efficiency and scalability info and see under for a abstract of efficiency throughout a various choice of broadly used excessive efficiency computing (HPC) workloads.
Figure 1: Performance comparability abstract of Azure HBv4/HX-series VMs to HBv3-series throughout numerous engineering and scientific computing workloads.
Azure HBv4 and HX-series VMs can be found in the present day within the Azure East United States area, and can quickly come to the Azure Korea Central, South Central US, Sweden Central, and Southeast Asia areas.
Faster, more economical, and extra energy environment friendly applied sciences for HPC
Azure HBv4 and HX sequence VMs characteristic 4th Gen AMD EPYC™ processors with AMD 3D V-cache™ as a result of they ship the quickest ranges of efficiency to a wide range of reminiscence performance-bound HPC workloads. The 2.3 GB L3 cache per VM can ship as much as 5.7 terabytes per second of bandwidth to amplify as much as 780 gigabytes per second of bandwidth from major reminiscence, for a market main blended common of 1.2 terabytes per second of efficient reminiscence bandwidth throughout a broad vary of buyer workloads. For widely-used reminiscence bandwidth-bound workloads like OpenFOAM, HBv4-series VMs with 4th Gen AMD EPYC CPUs with 3D V-Cache know-how are already yielding as much as 1.49 instances greater efficiency than normal 4th Gen EPYC processors in Azure’s inside testing.
Figure 2: Performance comparability of HBv4/HX throughout Preview (no 3D V-Cache) versus General Availability (consists of 3D V-Cache) from 1-8 VMs on CFD workload OpenFOAM.
Azure HBv4 and HX sequence VMs ship these vital efficiency enhancements at decrease price and vitality consumed per job straight proportional to the speed of workload speedup (corresponding to 25 p.c quicker efficiency on a given HPC workload with AMD 3D V-Cache in comparison with normal 4th Gen AMD EPYC CPUs interprets to the client incurring 25 p.c decrease price and energy consumed per job).
Learn extra about efficiency and scalability throughout a variety of functions, fashions, and configurations, and go to Azure Docs for full technical specs of HBv4 and HX sequence VMs.
Continuous enchancment for much more HPC prospects
Microsoft is delivering a brand new period of high-performance computing within the cloud; one outlined by steady enhancements to the crucial analysis and enterprise workloads that matter most to our prospects. Through our partnership with AMD, we’re making this imaginative and prescient a actuality by elevating the bar on the efficiency, scalability, and worth we ship with each launch of Azure H-series household VMs for our HPC prospects.
With the discharge of Azure HX sequence VMs, a brand new class inside the Azure H-series household, we’re extending our dedication to prospects operating giant reminiscence workloads like structural evaluation and semiconductor design. The 1.4 terabytes of reminiscence in each HX-series VM allows prospects to run extra of an information intensive workload out of reminiscence for vital speedups and price reductions. For broadly used automotive simulators like MSC Nastran, Azure’s inside testing reveals the brand new HX sequence VMs yield as much as 5.7 instances greater efficiency than HBv3-series VMs, and 9.2 instances greater efficiency than HC-series VMs launched 4 years in the past that includes know-how nonetheless broadly utilized in on-premises HPC environments in the present day.
Figure 3: Performance uplift from 2019–2023 on Azure H-series household Virtual Machines for structural evaluation software MSC NASTRAN.
Customer Momentum
“Materials sciences researchers, including those working with Azure Quantum, stand to benefit greatly from the introduction of Azure HBv4 virtual machines featuring powerful new processors and networking technologies. The 4x performance increase on NAMD shows once again Microsoft’s commitment to continuously making the most advanced computing resources available through Azure cloud services.”—Nathan Baker, Senior Director, Partnerships for Chemistry and Materials, Azure Quantum.
“AMD EPYC™ processors, available through the Azure HB-series family of Virtual Machines via Ansys Cloud, have been instrumental in advancing our ability to deliver more CFD simulations in support of our Motorsports efforts in North America. We firmly believe that accelerating our simulation capabilities and the engineering insights gained from this work have been key to our recent string of successes in both IndyCar and IMSA prototypes. We are looking forward to evaluating the performance of Azure HBv4 and HX-series VMs powered by the 4th Generation of AMD EPYC™ processors with AMD 3D V-Cache™ technology in the very near future. We are confident that AMD and Azure will continue to deliver on their promise of sustainable performance to be realized in time-to and cost-of solution for our very complex CFD models with these processors.”—Kelvin Fu, Vice President at Honda Performance Development.
“With Azure high-performance computing, we can run more jobs and tasks in parallel. We reduce the time to market by increasing the efficiency and the quality of the product that we release. This solution helps us to improve the product quality and reduce the risk of a delivery delay.”—Anna Palo, Digital Design Architect, STMicroelectronics.