VM sizes and types for Azure Stack Edge Pro

APPLIES TO: Yes for Pro GPU SKUAzure Stack Edge Pro - GPUYes for Pro 2 SKUAzure Stack Edge Pro 2Yes for Pro R SKUAzure Stack Edge Pro RYes for Mini R SKUAzure Stack Edge Mini R                  

This article describes the supported sizes for the virtual machines running on your Azure Stack Edge Pro devices. Use this article before you deploy virtual machines on your Azure Stack Edge Pro devices.

Supported VM sizes

The VM size determines the amount of compute resources (like CPU, GPU, and memory) that are made available to the VM. You should create virtual machines by using a VM size appropriate for the workload. Even though all machines will be running on the same hardware, machine sizes have different limits for disk access. This can help you manage overall disk access across your VMs. If a workload increases, you can also resize an existing virtual machine.

The following VMs are supported for creation on your Azure Stack Edge device.

Dv2-series

Size  vCPU  Memory (GiB)  Temp storage (GiB)  Max OS disk throughput (IOPS)  Max temp storage throughput (IOPS)  Max data disk throughput (IOPS)  Max data disks  Max NICs 
Standard_D1_v2  3.5  50  1000  3000  500 
Standard_D2_v2  100  1000  6000  500 
Standard_D3_v2  14  200  1000  12000  500  16 
Standard_D4_v2  28  400  1000  24000  500  32 
Standard_D5_v2  16  56  800  1000  48000  500  64 
Standard_D11_v2  14  100  1000  6000  500 
Standard_D12_v2  28  200  1000  12000  500  16 
Standard_D13_v2  56  400  1000  24000  500  32 

DSv2-series

Size  vCPU  Memory (GiB)  Temp storage (GiB)  Max OS disk throughput (IOPS)  Max temp storage throughput (IOPS)  Max data disk throughput (IOPS)  Max data disks  Max NICs 
Standard_DS1_v2  3.5  2000  4000  2300 
Standard_DS2_v2  14  2000  8000  2300 
Standard_DS3_v2  14  28  2000  16000  2300  16 
Standard_DS4_v2  28  56  2000  32000  2300  32 
Standard_DS5_v2  16  56  112  2000  64000  2300  64 
Standard_DS11_v2  14  28  2000  8000  2300 
Standard_DS12_v2  28  56  2000  16000  2300  16 
Standard_DS13_v2  56  112  2000  32000  2300  32 

For more information, see Dv2 and DSv2-series.

N-series GPU optimized

These sizes are supported for GPU VMs on your device and are optimized for compute-intensive GPU-accelerated applications, for example, inferencing workloads. The GPU VM that you deploy should match the GPU type on your Azure Stack Edge device.

NVIDIA Tesla T4 GPU

Size  vCPU  Memory (GiB)  Temp storage (GiB)  Max OS disk throughput (IOPS)  Max temp storage throughput (IOPS)  Max data disk throughput (IOPS)  GPU  GPU memory (GiB)  Max NICs 
Standard_NC4as_T4_v3  28  176  2000  48000  2300  16 
Standard_NC8as_T4_v3  56  352  2000  48000  2300  16  8
Standard_NC16as_T4_v3   16 110  352  2000  48000  2300  16  8

For more information, see NCasT4_v3-series.

NVIDIA A2 Tensor Core GPU

Size  vCPU  Memory (GiB)  Temp storage (GiB)  Max OS disk throughput (IOPS)  Max temp storage throughput (IOPS)  Max data disk throughput (IOPS)  GPU  GPU memory (GiB)  Max NICs 
Standard_NC4as_A2  28  176  2000  48000  2300  16 
Standard_NC8as_A2  56  352  2000  48000  2300  16  8
Standard_NC16as_A2   16 110  352  2000  48000  2300  16  8

F-series

These series are optimized for computational workloads and run on Intel Xeon processors.

Size  vCPU  Memory: GiB  Temp storage (GiB)  Max OS disk throughput (IOPS)  Max temp storage throughput (IOPS)  Max data disk throughput (IOPS)  Max data disks  Max NICs 
Standard_F1  16  1000  3000  500 
Standard_F2  32  1000  6000  500 
Standard_F4  64  1000  12000  500  16 
Standard_F8  16  128  1000  24000  500  32 
Standard_F12  12  24 256  1000  48000  500  64  8
Standard_F16  16  32  256  1000  48000  500  64 
Standard_F1s  2000  4000  2300 
Standard_F2s  2000  8000  2300 
Standard_F4s  16  2000  16000  2300  16 
Standard_F8s  16  32  2000  32000  2300  32 
Standard_F16s  16  32  64  2000  64000  2300  64  8
Standard_F4s_v1 4 4 8 2000 8000 2300 8 4

For more information, see Fsv2-series.

High-performance network VMs

The high-performance network (HPN) virtual machines are tailored for workloads that require fast and uninterrupted performance using high speed network interfaces. Due to the nature of logical core pairing, the supported VM sizes have vCpu count in multiples of 2.

HPN DSv2-series

Size  vCPU  Memory (GiB)  Temp storage (GiB)  Max OS disk throughput (IOPS)  Max temp storage throughput (IOPS)  Max data disk throughput (IOPS)  Max data disks  Max NICs1 
Standard_DS2_v2_HPN  14  2000  8000  2300  14 
Standard_DS3_v2_HPN  14  28  2000  16000  2300  16  14 
Standard_DS4_v2_HPN  28  56  2000  32000  2300  32  14 

1Windows Server 2016 Datacenter VHD has a limit of 8 NICs for all HPN VM sizes.

HPN F-series

Size  vCPU  Memory (GiB) Temp storage (GiB)  Max OS disk throughput (IOPS)  Max temp storage throughput (IOPS)  Max data disk throughput (IOPS)  Max data disks  Max NICs1 
Standard_F2s_HPN  2000  8000  2300  14 
Standard_F4s_HPN  16  2000  16000  2300  16  14 
Standard_F8s_HPN  16  32  2000  32000  2300  32  14 
Standard_F12s_HPN 12 24 48 2000 48000 2300 64 14
Standard_F16s_HPN  16  32  64  2000  64000  2300  64  14
Standard_F12_HPN  12 24  64 1000  48000  500  64  14 

1Windows Server 2016 Datacenter VHD has a limit of 8 NICs for all HPN VM sizes.

HPN and NVIDIA Tesla T4 GPU series

Size  vCPU  Memory (GiB) Temp storage (GiB)  Max OS disk throughput (IOPS)  Max temp storage throughput (IOPS)  Max data disk throughput (IOPS)  Max data disks  Max NICs1  GPU GPU memory (GiB)
Standard_DS2_v2_HPN_T4_v3  14  2000  8000  2300  14  1 16
Standard_DS3_v2_HPN_T4_v3  14  28  2000  16000  2300  16  14  1 16
Standard_DS4_v2_HPN_T4_v3  28  56  2000  32000  2300  32  14  1 16
Standard_F2s_HPN_T4_v3  2000  8000  2300  14  1 16
Standard_F4s_HPN_T4_v3  16  2000  16000  2300  16  14  1 16
Standard_F8s_HPN_T4_v3  16  32  2000  32000  2300  32  14  1 16
Standard_F12s_HPN_T4_v3  12  24  48  2000  48000  2300  64  14  1 16
Standard_F16s_HPN_T4_v3  16  32  64  2000  64000  2300  64  14  1 16
Standard_F12_HPN_T4_v3  12  24  64  1000  48000  500  64  14  1 16

1Windows Server 2016 Datacenter VHD has a limit of 8 NICs for all HPN VM sizes.

HPN and NVIDIA Tesla A2 Tensor Core GPU series

Size  vCPU  Memory (GiB) Temp storage (GiB)  Max OS disk throughput (IOPS)  Max temp storage throughput (IOPS)  Max data disk throughput (IOPS)  Max data disks  Max NICs1  GPU GPU memory (GiB)
Standard_DS2_v2_HPN_A2  14  2000  8000  2300  14  1 16
Standard_DS3_v2_HPN_A2  14  28  2000  16000  2300  16  14  1 16
Standard_DS4_v2_HPN_A2  28  56  2000  32000  2300  32  14  1 16
Standard_F2s_HPN_A2  2000  8000  2300  14  1 16
Standard_F4s_HPN_A2  16  2000  16000  2300  16  14  1 16
Standard_F8s_HPN_A2  16  32  2000  32000  2300  32  14  1 16
Standard_F12s_HPN_A2  12  24  48  2000  48000  2300  64  14  1 16
Standard_F16s_HPN_A2  16  32  64  2000  64000  2300  64  14  1 16
Standard_F12_HPN_A2  12  64  64  1000  48000  500  64  14  1 16

1Windows Server 2016 Datacenter VHD has a limit of 8 NICs for all HPN VM sizes.

Unsupported VM operations and cmdlets

Scale sets, availability sets, and snapshots aren't supported.

Next steps

Deploy VMs on your Azure Stack Edge Pro GPU device via the Azure portal