At Ignite 2019 this week, Microsoft’s Azure cloud team and AMD announced an expansion of their partnership that began in 2017 when Azure debuted Epyc-backed instances for storage workloads. The fourth-generation Azure D-series and E-series virtual machines previewed at the Rome launch in August are now generally available; they are powered by the Epyc “Rome” 7452 processors. Azure is also introducing the NVv4 instance series for virtual desktops, powered by Epyc 7742 CPUs and Radeon Instinct MI25 GPUs.
The latest Azure D-series, the Da_v4 and Das_v4 instances, target enterprise-grade applications, relational databases, in-memory caching, and analytics, and are positioned as the fastest Azure VMs in their class. Featuring the 32-core Epyc 7452 processors, these instances support up to 96 vCPUs, 384GB DDR4 RAM, and 2.4TB of SSD-based temporary storage per virtual machine.
The Microsoft Azure Ea_v4 and Eas_v4 VMs are optimized for large in-memory business critical workloads. Also based on the Eypc Rome 7452 parts, the E-Series can be spec’d with up to 96 vCPUs, up to 672GB DDR4 RAM, and 2.4TB SSD-based temporary storage per virtual machine. AMD and Microsoft claim the Ea-series VMs offer a 22 percent better performance/dollar than competitive VMs.
The distinction of the Das and Eas instances is that they offer Azure’s premium SSD managed disks, intended to support I/O intensive workloads with significantly high throughput and low latency.
All four of these Azure VMs are generally available now.
The new NVv4 instance, in preview mode, is powered by the 64-core AMD Epyc 7742 CPU and the AMD Vega MI25 GPU. It is “the first VM on Microsoft Azure to take advantage of SR-IOV technologies (Single-root input/output virtualization) and introduces GPU partitioning across four new options,” according to AMD’s George Watkins. With a single GPU’s resources supporting as many as eight virtual machines, NVv4 is intended as a cost-effective way to do visualization workloads on virtual Windows desktops.
For HPCers, Azure is previewing its HBv2 instance, based on the 7742 Epyc, the top “standard” Rome SKU (i.e., not including the higher-wattage 7H12 Epyc part). Announced in August at the Rome launch, HBv2 is the first cloud instance to feature 200 gigabit InfiniBand and can scale to 80,000 cores. AMD has indicated that the high-performance instances will be generally available by year end, first coming to the South Central US and West Europe Azure regions.
“With our new Azure HB and HC Virtual Machines, Azure is democratizing high-performance computing with unprecedented performance, scalability, and cost-efficiency for large tightly-coupled workloads in the cloud,” said Erin Chapple Corporate Vice President, Microsoft Azure Compute. Azure’s HC-series instances are based on Intel Xeon Skylake processors and leverage 100 Gbps InfiniBand from Mellanox.
The cloud company also reported that more than half of Azure workloads are running Linux.