ND A100 v4-series
Applies to: ✔️ Linux VMs ✔️ Flexible scale sets ✔️ Uniform scale sets
The ND A100 v4 series virtual machine is a new flagship addition to the Azure GPU family, designed for high-end Deep Learning training and tightly-coupled scale-up and scale-out HPC workloads.
The ND A100 v4 series starts with a single virtual machine (VM) and eight NVIDIA Ampere A100 40GB Tensor Core GPUs. ND A100 v4-based deployments can scale up to thousands of GPUs with an 1.6 Tb/s of interconnect bandwidth per VM. Each GPU within the VM is provided with its own dedicated, topology-agnostic 200 Gb/s NVIDIA Mellanox HDR InfiniBand connection. These connections are automatically configured between VMs occupying the same virtual machine scale set, and support GPUDirect RDMA.
Each GPU features NVLINK 3.0 connectivity for communication within the VM, and the instance is also backed by 96 physical 2nd-generation AMD Epyc™ 7V12 (Rome) CPU cores.
These instances provide excellent performance for many AI, ML, and analytics tools that support GPU acceleration 'out-of-the-box,' such as TensorFlow, Pytorch, Caffe, RAPIDS, and other frameworks. Additionally, the scale-out InfiniBand interconnect is supported by a large set of existing AI and HPC tools built on NVIDIA's NCCL2 communication libraries for seamless clustering of GPUs.
To get started with ND A100 v4 VMs, refer to HPC Workload Configuration and Optimization for steps including driver and network configuration. Due to increased GPU memory I/O footprint, the ND A100 v4 requires the use of Generation 2 VMs and marketplace images. The Azure HPC images are strongly recommended. Azure HPC Ubuntu 18.04, 20.04 and Azure HPC CentOS 7.9 images are supported.
Premium Storage: Supported
Premium Storage caching: Supported
Ultra Disks: Supported (Learn more about availability, usage, and performance)
Live Migration: Not Supported
Memory Preserving Updates: Not Supported
VM Generation Support: Generation 2
Accelerated Networking: Supported
Ephemeral OS Disks: Supported
InfiniBand: Supported, GPUDirect RDMA, 8 x 200 Gigabit HDR
Nvidia NVLink Interconnect: Supported
Nested Virtualization: Not Supported
The ND A100 v4 series supports the following kernel versions:
CentOS 7.9 HPC: 3.10.0-1160.24.1.el7.x86_64
Ubuntu 18.04: 5.4.0-1043-azure
Ubuntu 20.04: 5.4.0-1046-azure
|Size||vCPU||Memory: GiB||Temp Storage (SSD): GiB||GPU||GPU Memory: GiB||Max data disks||Max uncached disk throughput: IOPS / MBps||Max network bandwidth||Max NICs|
|Standard_ND96asr_v4||96||900||6000||8 A100 40 GB GPUs (NVLink 3.0)||40||32||80,000 / 800||24,000 Mbps||8|
Size table definitions
Storage capacity is shown in units of GiB or 1024^3 bytes. When you compare disks measured in GB (1000^3 bytes) to disks measured in GiB (1024^3) remember that capacity numbers given in GiB may appear smaller. For example, 1023 GiB = 1098.4 GB.
Disk throughput is measured in input/output operations per second (IOPS) and MBps where MBps = 10^6 bytes/sec.
Data disks can operate in cached or uncached modes. For cached data disk operation, the host cache mode is set to ReadOnly or ReadWrite. For uncached data disk operation, the host cache mode is set to None.
To learn how to get the best storage performance for your VMs, see Virtual machine and disk performance.
Expected network bandwidth is the maximum aggregated bandwidth allocated per VM type across all NICs, for all destinations. For more information, see Virtual machine network bandwidth.
Upper limits aren't guaranteed. Limits offer guidance for selecting the right VM type for the intended application. Actual network performance will depend on several factors including network congestion, application loads, and network settings. For information on optimizing network throughput, see Optimize network throughput for Azure virtual machines. To achieve the expected network performance on Linux or Windows, you may need to select a specific version or optimize your VM. For more information, see Bandwidth/Throughput testing (NTTTCP).
Other sizes and information
- General purpose
- Memory optimized
- Storage optimized
- GPU optimized
- High performance compute
- Previous generations
Pricing Calculator : Pricing Calculator
For more information on disk types, see What disk types are available in Azure?
Learn more about how Azure compute units (ACU) can help you compare compute performance across Azure SKUs.