Microsoft opened a public preview of the new Mbv4-series Azure virtual machines on Tuesday, bringing substantially higher NVMe storage performance to memory-intensive enterprise workloads like SAP HANA. The VMs are available now in the East US and East US 2 regions, with broader rollout expected later this year.

This is the first time an Azure VM family pairs the latest AMD EPYC processors with purpose-built NVMe caching to slash database response times. For IT admins running mission-critical SAP landscapes, that means lower latency on critical queries and smoother operations during peak load without jumping to a higher instance size.

What the Mbv4 Series Actually Delivers

Mbv4 is Microsoft’s fourth-generation memory-optimized virtual machine family, designed for workloads that demand the highest ratio of RAM to vCPU. Every instance in the family includes at least 8 GiB of memory per vCPU, but the real headline is the storage.

Instead of relying solely on the same remote Premium SSD options used by earlier M-series VMs, Mbv4 includes a local NVMe cache engine that accelerates both read and write operations. Microsoft’s early benchmarks show up to a 3.2x improvement in throughput for 8 KB random reads compared with the current Mbv3 series. That translates directly into faster SAP HANA data loads, quicker analytical queries, and tighter batch windows for enterprises.

The preliminary instance lineup includes four sizes:

  • Standard_Mb4_v4 – 4 vCPUs, 32 GiB RAM, 400 GiB local NVMe
  • Standard_Mb8_v4 – 8 vCPUs, 64 GiB RAM, 600 GiB local NVMe
  • Standard_Mb16_v4 – 16 vCPUs, 128 GiB RAM, 900 GiB local NVMe
  • Standard_Mb32_v4 – 32 vCPUs, 256 GiB RAM, 1.8 TiB local NVMe

All sizes ship with Azure Boost, a pre-configured acceleration stack that offloads networking and storage I/O onto dedicated hardware. That means fewer CPU cycles wasted on I/O handling and more compute headroom for your application. The preview supports both Windows Server 2022 and select Linux distributions certified for SAP, including SUSE Linux Enterprise Server 15 SP5 and Red Hat Enterprise Linux 9.2.

What It Means for You

For SAP Basis and database administrators: If you’re running an SAP landscape on Azure, the jump in NVMe throughput is the kind of improvement that can postpone a hardware refresh. By moving from Mbv3 to Mbv4, you may be able to stay on the same instance size while seeing significantly better database performance. Early adopters in the preview report that SAP HANA startup times dropped by as much as 40 percent on comparable instance sizes thanks to the faster local storage.

One nuance: the local NVMe is ephemeral – it does not persist across reboots or deallocations. That’s perfect for SAP HANA’s data volume caching or for TempDB in SQL Server, but you’ll still attach Azure Premium SSD or Ultra Disk for your persistent data and log volumes. Microsoft recommends placing the HANA data files on premium managed disks and using the NVMe purely for the log buffer and shared memory segments.

For IT managers and architects: This preview arrives just as many organizations are planning their next Azure migration wave or capacity update for SAP S/4HANA. The new VMs may change your sizing calculations – a smaller Mbv4 instance could match or exceed the throughput of a larger Mbv3 instance, potentially lowering monthly compute cost. However, because preview instances typically do not carry an SLA, you’ll want to keep your production workloads on generally available SKUs for now.

For developers and test environments: The Mbv4 preview is an excellent opportunity to benchmark your exact workload. You can spin up a test SAP system, run your critical ABAP reports, and compare execution times head-to-head with your current instances. Just be mindful that preview regions are limited to East US and East US 2, so latency from your on-premises networks may skew results if you’re testing from outside the US East Coast.

How We Got Here

Microsoft’s memory-optimized VM lineage traces back to the original M-series in 2016, built on Intel Xeon processors. Those VMs delivered up to 4 TB of RAM but relied entirely on remote storage, creating a bottleneck for in-memory databases that also needed fast caching. With the Mbv3 series in 2022, Microsoft introduced AMD EPYC processors and Azure Boost, but local storage was still based on slower SATA SSDs or no local disk at all in some sizes.

The jump to Mbv4 reflects a broader industry trend: SAP HANA certifications increasingly require specific NVMe performance thresholds, and hyperscalers are responding with purpose-built hardware. AWS tackled this with its X2iedn instances last year, and Google Cloud followed with its M3 series. Microsoft’s playbook has been to match or exceed those baseline specs while leaning on Azure Boost to reduce the I/O tax on CPUs.

Coupled with that, the East US 2 region has become a proving ground for AMD EPYC-based instances within Azure. Milan-X processors with 3D V-Cache are likely under the hood here, based on Microsoft’s teaser benchmarks that show dramatic improvements in SQL Server and SAP HANA workloads that benefit from larger L3 caches.

What to Do Now

If you want to test Mbv4 for your SAP or memory-heavy application:

  1. Check region availability. The preview is live in East US and East US 2 zones. You can deploy instances immediately through the Azure portal, CLI, or ARM templates. Select the “Mbv4” family when sizing your VM.
  2. Request quota increase if needed. New VM families often have default core quotas of zero. Go to the Azure subscription’s “Usage + quotas” blade and request an increase for the “Standard Mbv4 Family vCPUs.” During public preview, Microsoft may impose temporary limits of no more than 64 vCPUs per subscription.
  3. Use a certified operating system image. For SAP workloads, pick an image from the Azure Marketplace that carries the “SAP” badge. Avoid using Windows Server 2019; it lacks the updated NVMe drivers that ship with Server 2022 and may not recognize the local disk correctly.
  4. Configure storage correctly. In your deployment template, attach a premium SSD for the OS disk, one or more Ultra Disks for SAP data, and then configure the SAP HANA storage layout to place the global.ini and trace files on the ephemeral NVMe. Microsoft’s documentation includes sample scripts to automate this.
  5. Monitor with Azure Boost diagnostics. Because Azure Boost handles the NVMe I/O, you’ll find new performance counters under the "/Azure Boost" namespace in Azure Monitor. Set up a dashboard to track latency and IOPS during your test runs.

Take note: preview pricing is typically 30–40% lower than the eventual GA price, but instances are not covered by the Azure SLA. Don’t move production systems onto Mbv4 until the series goes generally available and SAP certification is finalized.

What to Watch Next

The Mbv4 preview will likely run through the end of Q3 2025, with general availability expected in October. Two factors will determine how quickly enterprises adopt the new series: SAP certification for the exact hardware configuration and the availability of Mbv4 in more regions – specifically West Europe and Southeast Asia, where large SAP landscapes commonly run. Microsoft has hinted that Mbv4 will also support nested virtualization for AI workloads, though that feature may slip to a post-GA update. In the meantime, early testers in the East US regions are getting a first look at the fastest memory-optimized VMs Azure has ever released.