Infrastructure Support Matrix - Release 7.4

Use this support matrix to verify your infrastructure is compatible with NVIDIA AI Enterprise release 7.4. This page provides comprehensive compatibility information for GPUs, operating systems, hypervisors, orchestration platforms, and cloud providers.

Note

For the latest support matrix information and to compare across multiple releases (7.0 - 7.4), use the Interactive Support Matrix.

How to Use This Matrix

1. Identify your deployment type - On-premises (bare metal or virtualized) or public cloud

2. Verify your GPU - Check the Supported GPUs section

3. Confirm your platform - Match your OS, hypervisor, and orchestration platform

4. Check software versions - Ensure your infrastructure software versions are supported

Quick Navigation

• Supported GPUs and Networking - GPU and network card compatibility

• Bare Metal Deployments - Kubernetes and standalone containers on physical servers

• Virtualized Deployments - Hypervisor and guest OS combinations

• Public Cloud - Managed Kubernetes and standard GPU instances

• Base Command Manager - Cluster management platform support

Requirements

For NVIDIA AI Enterprise to be supported, ensure you meet the following requirements.

On-Premises Servers

1. GPU supported by NVIDIA AI Enterprise (and optionally a supported networking card)

2. NVIDIA-Certified Systems

3. Supported infrastructure management software - bare-metal or virtualized (OS, orchestration platform, hypervisor)

Cloud Servers

1. Supported orchestration platform

2. Supported virtual machine instance - standard or optimized

Supported NVIDIA Infrastructure Software

Table 6 Supported Infrastructure Software

Product	Version	x86	ARM	Government Ready
Core Infrastructure Drivers
NVIDIA GPU Data Center Driver	580.126.09	Supported	Supported	N/A
NVIDIA Fabric Manager	580.126.09	Supported	Supported	N/A
NVIDIA DOCA-OFED Driver for Networking	3.2.0	Supported	Supported	Supported
Virtualization
NVIDIA vGPU for Compute (Virtual GPU Manager and Guest Drivers)	19.4	Supported	N/A	N/A
Container Platform
NVIDIA Container Toolkit	1.18.1	Supported	Supported	N/A
Kubernetes Operators
NVIDIA GPU Operator	25.10.1	Supported	Supported (N/A for Government Ready)	Supported [11]
NVIDIA Network Operator	25.10.0	Supported	Supported	Supported [12]
NVIDIA DPU Operator (DPF)	25.10.1	Supported	N/A	N/A
NVIDIA NIM Operator	3.0.2	Supported	Supported	Supported
Data Center Services
NVIDIA DOCA Microservices	3.2.1	Supported	Supported	N/A
Cluster Management & Orchestration
NVIDIA Base Command Manager (BCM)	11.31.0	Supported	Supported	N/A

Supported NVIDIA GPUs and Networking

NVIDIA AI Enterprise is supported on the following NVIDIA GPUs with compatible third-party servers listed on the NVIDIA-Certified Systems page.

Specific NVIDIA AI Enterprise-supported products may not support all operating systems or GPUs. Refer to the individual product release notes for discrepancies.

Note

For HGX support information, refer to the Supported Platforms section.

NVIDIA Blackwell

• NVIDIA RTX Pro 6000 Blackwell Server Edition

NVIDIA Hopper

• NVIDIA H800

• NVIDIA H800 NVL

• NVIDIA H200 NVL

• NVIDIA H100

• NVIDIA H100 NVL

• NVIDIA H20

NVIDIA Ada Lovelace

• NVIDIA L40S

• NVIDIA L40

• NVIDIA L20

• NVIDIA L4

• NVIDIA L2

• NVIDIA RTX 6000 Ada Generation

• NVIDIA RTX 5880 Ada Generation

• NVIDIA RTX 5000 Ada Generation

• NVIDIA RTX 4000 SFF Ada Generation

NVIDIA Ampere

• NVIDIA A800

• NVIDIA AX800

• NVIDIA A100X

• NVIDIA A100

• NVIDIA A40

• NVIDIA A30X

• NVIDIA A30

• NVIDIA A16

• NVIDIA A10

• NVIDIA A10G

• NVIDIA A10M

• NVIDIA A2

• NVIDIA RTX A6000

• NVIDIA RTX A5000

• NVIDIA RTX A4000

NVIDIA Turing

• NVIDIA T4

• NVIDIA T4G

• NVIDIA Quadro RTX 8000

• NVIDIA Quadro RTX 6000

• NVIDIA Quadro RTX 4000

NVIDIA Volta

• NVIDIA V100

Note

Multiple GPU architectures can be deployed within the same NVIDIA AI Enterprise environment.

Multi-node requires an Ethernet NIC that supports RoCE. NVIDIA recommends using an NVIDIA Mellanox ConnectX and an NVIDIA GPU.

Table 7 Supported Ethernet NICs and SuperNICs

Product Family	Architecture
NVIDIA ConnectX-6 NIC	NVIDIA ConnectX-6
NVIDIA ConnectX-6 Dx NIC	NVIDIA ConnectX-6 Dx
NVIDIA ConnectX-7 NIC	NVIDIA ConnectX-7
NVIDIA ConnectX-8 NIC	NVIDIA ConnectX-8
NVIDIA BlueField-3 SuperNIC	NVIDIA BlueField-3
NVIDIA BlueField-3 DPU	NVIDIA BlueField-3

Supported Platforms

NVIDIA AI Enterprise is supported on NVIDIA DGX servers in bare-metal deployments with the NVIDIA data center driver for Linux included in the DGX OS software.

Note

DGX platforms, HGX platforms with KVM hypervisors, and IGX Orin are not supported with NVIDIA vGPU for Compute. For NVIDIA IGX, refer to the NVIDIA AI Enterprise - IGX Packaging, Pricing, and Licensing Guide.

Table 8 Supported Platforms

Accelerated Platform	Architecture
NVIDIA DGX B300	NVIDIA Blackwell Ultra
NVIDIA HGX B300 [1] [14] [15]	NVIDIA Blackwell Ultra
NVIDIA DGX GB300 NVL72 [5]	NVIDIA Grace Blackwell
NVIDIA GB300 NVL72 [5]	NVIDIA Grace Blackwell
NVIDIA DGX B200	NVIDIA Blackwell
NVIDIA HGX B200 [1] [13] [14] [15]	NVIDIA Blackwell
NVIDIA DGX GB200 NVL72 [5]	NVIDIA Grace Blackwell
NVIDIA GB200 NVL72 [5]	NVIDIA Grace Blackwell
NVIDIA GH200 [5]	NVIDIA Grace Hopper
NVIDIA DGX H100	NVIDIA Hopper
NVIDIA HGX H100	NVIDIA Hopper
NVIDIA HGX H800	NVIDIA Hopper
NVIDIA HGX H200	NVIDIA Hopper
NVIDIA HGX H20	NVIDIA Hopper
NVIDIA IGX Orin [2]	NVIDIA Ada Lovelace
NVIDIA DGX A100	NVIDIA Ampere
NVIDIA HGX A100	NVIDIA Ampere
NVIDIA HGX A800	NVIDIA Ampere

Bare Metal Deployments

Use the following tables to verify compatibility for NVIDIA AI Enterprise on dedicated physical servers.

What’s Covered

• Kubernetes-orchestrated containers with supported distributions (Charmed, Red Hat OpenShift, SUSE RKE2, Upstream Kubernetes)

• Standalone Docker/Podman containers on Red Hat Enterprise Linux, SUSE Linux Enterprise Server, and Ubuntu

• GPU Operator, Network Operator, and DPU Operator support by platform

Table 9 Containers Orchestrated with Kubernetes on Bare Metal

Orchestration Platform			Operating System		NVIDIA AI Enterprise Infrastructure Support
Name	Versions	Engine	Name	Versions	GPU Operator	Network Operator	DPU Operator (DPF)	GPU Driver Support [3]
Charmed Kubernetes	1.30 - 1.34	Containerd	Ubuntu	20.04 LTS, 22.04 LTS, 24.04 LTS	Supported	Supported	N/A	vGPU Guest/Data Center
HPE Ezmeral Runtime Enterprise	5.6	Containerd	Red Hat Enterprise Linux	8.6, 8.8, 8.10, 9.4, 9.6	Supported	N/A	N/A	vGPU Guest/Data Center
Nutanix NKP	2.12, 2.13	Containerd	Ubuntu	20.04 LTS, 22.04 LTS, 24.04 LTS	Supported	Supported	N/A	vGPU Guest/Data Center
Red Hat OpenShift [4] [8]	4.16 - 4.20	CRI-O	Red Hat CoreOS	4.16 - 4.20	Supported	Supported [12]	N/A	vGPU Guest/Data Center
SUSE RKE2	1.30 - 1.34	Containerd	Ubuntu	20.04 LTS, 22.04 LTS, 24.04 LTS	Supported	Supported	N/A	vGPU Guest/Data Center
SUSE RKE2	1.30 - 1.34	Containerd	Red Hat Enterprise Linux	8.6, 8.8, 8.10, 9.4, 9.6	Supported	Supported	N/A	vGPU Guest/Data Center
Upstream Kubernetes	1.30 - 1.34	Containerd	Red Hat Enterprise Linux	8.6, 8.8, 8.10, 9.4, 9.6	Supported	Supported	Supported	vGPU Guest/Data Center
Upstream Kubernetes	1.30 - 1.34	Containerd	Ubuntu	20.04 LTS, 22.04 LTS, 24.04 LTS	Supported	Supported	Supported	vGPU Guest/Data Center
Upstream Kubernetes	1.30 - 1.34	CRI-O	Red Hat Enterprise Linux	8.6, 8.8, 8.10, 9.4, 9.6	Supported	Supported	Supported	vGPU Guest/Data Center
Upstream Kubernetes	1.30 - 1.34	CRI-O	Ubuntu	20.04 LTS, 22.04 LTS, 24.04 LTS	Supported	N/A	N/A	vGPU Guest/Data Center

Table 10 Standalone Containers on Bare Metal

Container		Operating System		NVIDIA AI Enterprise Infrastructure Support
Name	Engine	Name	Versions	GPU Operator	Network Operator	DPU Operator (DPF)	GPU Driver Support [3]
Non-Kubernetes (standalone containers)	Docker/Podman	Red Hat Enterprise Linux	7.9, 8.6, 8.8, 8.10, 9.2, 9.4, 9.6	N/A	N/A	N/A	vGPU Guest/Data Center
Non-Kubernetes (standalone containers)	Docker/Podman	SUSE Linux Enterprise Server	15 SP2 and later	N/A	N/A	N/A	vGPU Guest/Data Center
Non-Kubernetes (standalone containers)	Docker/Podman	Ubuntu	20.04 LTS, 22.04 LTS, 24.04 LTS	N/A	N/A	N/A	vGPU Guest/Data Center

Virtualized Deployments

Use the following tables to verify compatibility for NVIDIA AI Enterprise in virtualized environments with NVIDIA vGPU for Compute.

What’s Covered

• Kubernetes-orchestrated containers on VMware vSphere, Red Hat KVM, Ubuntu KVM, and Nutanix AHV

• Standalone Docker/Podman containers with supported hypervisor and guest OS combinations

• Non-containerized applications on supported hypervisors with vGPU

• Both vGPU and GPU passthrough modes (where applicable)

Table 11 Containers Orchestrated with Kubernetes on Virtualized Environments

Orchestration Platform Name	Versions	Engine	Guest OS Name	Guest OS Versions	Hypervisor Name	Hypervisor Versions	GPU Operator	Network Operator	DPU Operator (DPF)	GPU Driver Support (vGPU)	GPU Driver Support (Passthrough) [6]
Charmed Kubernetes	1.30 - 1.34	Containerd	Ubuntu	20.04 LTS, 22.04 LTS, 24.04 LTS	Ubuntu [7] [10]	20.04 LTS, 22.04 LTS, 24.04 LTS	Supported	Supported	N/A	vGPU Guest	vGPU Guest/Data Center
Charmed Kubernetes	1.30 - 1.34	Containerd	Ubuntu	20.04 LTS, 22.04 LTS, 24.04 LTS	VMware vSphere	8.0 and later, 9.0	Supported	Supported	N/A	vGPU Guest	vGPU Guest/Data Center
Nutanix NKP	2.12, 2.13	Containerd	Ubuntu	20.04 LTS, 22.04 LTS, 24.04 LTS	Nutanix AOS/AHV [7] [9] [10]	6.5 - 6.10	Supported	Supported	N/A	vGPU Guest	vGPU Guest/Data Center
Red Hat OpenShift [8]	4.16 - 4.20	CRI-O	Red Hat CoreOS	4.16 - 4.20	Red Hat Enterprise Linux [7] [10]	8.10, 9.4, 9.6	Supported	Supported [12]	N/A	vGPU Guest	vGPU Guest/Data Center
Red Hat OpenShift [8]	4.16 - 4.20	CRI-O	Red Hat CoreOS	4.16 - 4.20	VMware vSphere	8.0 and later, 9.0	Supported	Supported [12]	N/A	vGPU Guest	vGPU Guest/Data Center
Upstream Kubernetes	1.30 - 1.34	Containerd, CRI-O	Red Hat Enterprise Linux	8.6, 8.8, 8.10, 9.4, 9.6	Red Hat Enterprise Linux [7] [10]	8.10, 9.4, 9.6	Supported	Supported	N/A	vGPU Guest	vGPU Guest/Data Center
Upstream Kubernetes	1.30 - 1.34	Containerd, CRI-O	Red Hat Enterprise Linux	8.6, 8.8, 8.10, 9.4, 9.6	VMware vSphere	8.0 and later, 9.0	Supported	Supported	N/A	vGPU Guest	vGPU Guest/Data Center
Upstream Kubernetes	1.30 - 1.34	Containerd, CRI-O	Red Hat Enterprise Linux	8.6, 8.8, 8.10, 9.4, 9.6	Nutanix AOS/AHV [7] [9] [10]	6.5 - 6.10	Supported	Supported	N/A	vGPU Guest	vGPU Guest/Data Center
Upstream Kubernetes	1.30 - 1.34	Containerd, CRI-O	Ubuntu	20.04 LTS, 22.04 LTS, 24.04 LTS	Ubuntu [7] [10]	20.04 LTS, 22.04 LTS, 24.04 LTS	Supported	Supported	N/A	vGPU Guest	vGPU Guest/Data Center
Upstream Kubernetes	1.30 - 1.34	Containerd, CRI-O	Ubuntu	20.04 LTS, 22.04 LTS, 24.04 LTS	VMware vSphere	8.0 and later, 9.0	Supported	Supported	N/A	vGPU Guest	vGPU Guest/Data Center
Upstream Kubernetes	1.30 - 1.34	Containerd, CRI-O	Ubuntu	20.04 LTS, 22.04 LTS, 24.04 LTS	Nutanix AOS/AHV [7] [9] [10]	6.5 - 6.10	Supported	Supported	N/A	vGPU Guest	vGPU Guest/Data Center
VMware vSphere Kubernetes Service	VMware K8 1.30 - 1.34	Containerd	Ubuntu	20.04 LTS, 22.04 LTS, 24.04 LTS	VMware vSphere	8.0 and later, 9.0	Supported	N/A	N/A	vGPU Guest	vGPU Guest/Data Center

Table 12 Standalone Containers on Virtualized Environments

Container Name	Engine	Guest OS Name	Guest OS Versions	Hypervisor Name	Hypervisor Versions	GPU Operator	Network Operator	DPU Operator (DPF)	GPU Driver Support (vGPU)	GPU Driver Support (Passthrough) [6]
Non-Kubernetes (standalone containers)	Docker/Podman	Red Hat Enterprise Linux	8.10, 9.4, 9.6, 9.7, 10.0, 10.1	Red Hat Enterprise Linux [7]	8.10, 9.4, 9.6, 9.7, 10.0, 10.1	N/A	N/A	N/A	vGPU Guest	vGPU Guest/Data Center
Non-Kubernetes (standalone containers)	Docker/Podman	Red Hat Enterprise Linux	8.10, 9.4, 9.6, 9.7	VMware vSphere	8.0 and later, 9.0	N/A	N/A	N/A	vGPU Guest	vGPU Guest/Data Center
Non-Kubernetes (standalone containers)	Docker/Podman	SUSE Linux Enterprise Server	15 SP2 and later	VMWare vSphere	8.0 and later, 9.0	N/A	N/A	N/A	vGPU Guest	vGPU Guest/Data Center
Non-Kubernetes (standalone containers)	Docker/Podman	Ubuntu	20.04 LTS, 22.04 LTS, 24.04 LTS	Ubuntu [7] [10]	20.04 LTS, 22.04 LTS, 24.04 LTS	N/A	N/A	N/A	vGPU Guest	vGPU Guest/Data Center
Non-Kubernetes (standalone containers)	Docker/Podman	Ubuntu	20.04 LTS, 22.04 LTS, 24.04 LTS	VMware vSphere	8.0 and later, 9.0	N/A	N/A	N/A	vGPU Guest	vGPU Guest/Data Center
Non-Kubernetes (standalone containers)	Docker/Podman	Ubuntu	20.04 LTS, 22.04 LTS, 24.04 LTS	Nutanix AOS/AHV [7] [9] [10]	6.5 - 6.10	N/A	N/A	N/A	vGPU Guest	vGPU Guest/Data Center

Table 13 Non-containerized Applications on Hypervisors and Guest Operating Systems Supported with vGPU

Hypervisor Name	Hypervisor Versions	Guest OS Name	Guest OS Versions	vGPU Driver Support
VMware ESX	ESXi 8.0 and later, 9.0	Debian	12	vGPU Guest
VMware ESX	ESXi 8.0 and later, 9.0	Red Hat Enterprise Linux	8.10, 9.4, 9.6, 9.7	vGPU Guest
VMware ESX	ESXi 8.0 and later, 9.0	Ubuntu	20.04 LTS, 22.04 LTS, 24.04 LTS	vGPU Guest
VMware ESX	ESXi 8.0 and later, 9.0	SUSE Linux Enterprise Server	12 SP3+, 12 SP5, 15 SP2	vGPU Guest
VMware ESX	ESXi 8.0 and later, 9.0	Microsoft Windows	Server 2022, Windows 10, Windows 11	vGPU Guest
Ubuntu [7] [10]	20.04 LTS, 22.04 LTS, 24.04 LTS	Ubuntu	20.04 LTS, 22.04 LTS, 24.04 LTS	vGPU Guest
Red Hat Enterprise Linux (RHEL) with KVM [7] [10]	8.10, 9.4, 9.6, 9.7, 10.0, 10.1	Red Hat Enterprise Linux	8.10, 9.4, 9.6, 9.7, 10.0, 10.1	vGPU Guest
Red Hat Enterprise Linux (RHEL) with KVM [7] [10]	8.10, 9.4, 9.6, 9.7, 10.0, 10.1	Microsoft Windows	Server 2022	vGPU Guest
Red Hat Enterprise Linux (RHEL) with KVM [7] [10]	Red Hat OpenStack Platform, Red Hat OpenStack Services on OpenShift	Product Documentation for Red Hat OpenStack	Product Documentation for Red Hat OpenStack	vGPU Guest
Nutanix AOS/AHV [7] [9] [10]	6.5 - 6.10	Ubuntu	20.04 LTS, 22.04 LTS, 24.04 LTS	vGPU Guest
Nutanix AOS/AHV [7] [9] [10]	6.5 - 6.10	Debian	12	vGPU Guest
Nutanix AOS/AHV [7] [9] [10]	6.5 - 6.10	Red Hat Enterprise Linux	8.8, 8.10, 9.2, 9.4, 9.6	vGPU Guest
Nutanix AOS/AHV [7] [9] [10]	6.5 - 6.10	SUSE Linux Enterprise Server	12, 15	vGPU Guest

Base Command Manager

NVIDIA Base Command Manager provides cluster provisioning, workload management, and infrastructure monitoring for data centers.

Supported Configurations

• Kubernetes orchestration

• Slurm workload manager (non-Kubernetes)

• Red Hat Enterprise Linux and Ubuntu operating systems

Table 14 Base Command Manager

Orchestration Platform Name	Versions	Engine	OS Name	OS Versions	GPU Operator	Network Operator	DPU Operator (DPF)	GPU Driver Support
Upstream Kubernetes	1.31 - 1.34	Containerd	Ubuntu	20.04 LTS, 22.04 LTS, 24.04 LTS	Supported	Supported	N/A	Data Center
Upstream Kubernetes	1.31 - 1.34	Containerd	Red Hat Enterprise Linux	8, 9	Supported	Supported	N/A	Data Center
Slurm (non-Kubernetes)	24.05, 24.11, 25.05	N/A	Ubuntu	20.04 LTS, 22.04 LTS, 24.04 LTS	N/A	N/A	N/A	Data Center
Slurm (non-Kubernetes)	24.05, 24.11, 25.05	N/A	Red Hat Enterprise Linux	8, 9	N/A	N/A	N/A	Data Center

Public Cloud

Managed Kubernetes

Verify compatibility for NVIDIA AI Enterprise on cloud-managed Kubernetes services.

Supported Cloud Providers

• Amazon Web Services (EKS)

• Google Cloud Platform (GKE)

• Microsoft Azure (AKS)

• Red Hat OpenShift (Managed Service - multi-cloud)

Table 15 Managed Kubernetes

Cloud Service Provider	Orchestration Platform Name	K8s Versions	Engine	OS Name	OS Versions	GPU Operator	Network Operator	DPU Operator (DPF)	GPU Driver Support
AWS	Amazon Elastic Kubernetes Service (EKS)	1.30 - 1.34	Containerd	Ubuntu	20.04 LTS, 22.04 LTS, 24.04 LTS	Supported	N/A	N/A	vGPU Guest/Data Center
Google	Google Kubernetes Engine (GKE)	1.30 - 1.34	Containerd	Ubuntu	22.04 LTS, 24.04 LTS	Supported	N/A	N/A	vGPU Guest/Data Center
Microsoft	Azure Kubernetes Service (AKS)	1.30 - 1.34	Containerd	Ubuntu	22.04 LTS, 24.04 LTS	Supported	N/A	N/A	vGPU Guest/Data Center
N/A	Red Hat OpenShift (Managed Service)	4.16 - 4.20	CRI-O	Red Hat CoreOS	4.16 - 4.20	Supported	N/A	N/A	vGPU Guest/Data Center

Standard GPU Instances

Standard GPU virtual machine instances across major cloud providers support NVIDIA AI Enterprise for both Kubernetes and standalone containers.

Table 16 Standard Instances for Kubernetes and Standalone Containers

Cloud Service Provider	Virtual Machine (VM) Instance with GPU	Product Family
Alibaba Cloud	gn7e, gn7i	NVIDIA A10
Alibaba Cloud	gn7s	NVIDIA A30
Alibaba Cloud	gn6i	NVIDIA T4
Alibaba Cloud	gn6e, gn6v	NVIDIA V100
Alibaba Cloud	ecs.ebmgn8v, ecs.gn8v	NVIDIA H20
Alibaba Cloud	ecs.ebmgn8is, ecs.gn8is	NVIDIA L20
Amazon Web Services	EC2 P3	NVIDIA V100
Amazon Web Services	EC2 P4	NVIDIA A100
Amazon Web Services	EC2 P5	NVIDIA H100
Amazon Web Services	EC2 P5e and P5en	NVIDIA H200
Amazon Web Services	EC2 G4	NVIDIA T4
Amazon Web Services	EC2 G5	NVIDIA A10G
Amazon Web Services	EC2 G6	NVIDIA L4
Amazon Web Services	EC2 G6e	NVIDIA L40S
Microsoft Azure	ND GB200-v6	NVIDIA GB200
Microsoft Azure	ND-H200-v5	NVIDIA H200
Microsoft Azure	ND-H100-v5, NCads_H100_v5-series, NCCads_H100_v5-series	NVIDIA H100
Microsoft Azure	NCv3-series	NVIDIA V100
Microsoft Azure	NCasT4_v3-series	NVIDIA T4
Microsoft Azure	NC_A100_v4-series	NVIDIA A100
Google Cloud Platform	A4 VM	NVIDIA B200
Google Cloud Platform	A3 VM	NVIDIA H100, H200
Google Cloud Platform	A2 VM	NVIDIA A100
Google Cloud Platform	G2 VM	NVIDIA L4
Google Cloud Platform	N1 VM	NVIDIA T4, V100
Oracle Cloud Infrastructure	BM.GPU3, VM.GPU3	NVIDIA V100
Oracle Cloud Infrastructure	BM.GPU4, BM.GPU.A100	NVIDIA A100
Oracle Cloud Infrastructure	BM.GPU.A10, VM.GPU.A10	NVIDIA A10
Oracle Cloud Infrastructure	BM.GPU.H100.8	NVIDIA H100
Tencent Cloud	PNV4	NVIDIA A10
Tencent Cloud	GT4	NVIDIA A100
Tencent Cloud	GN10Xp, GN10X	NVIDIA V100
Tencent Cloud	GN7, GN7vi, GI3X	NVIDIA T4
Volcano Engine	ecs.gni2	NVIDIA A10

NVIDIA GPU Optimized VMI on CSP Marketplace

NVIDIA provides pre-configured, validated Virtual Machine Instances (VMI) for rapid AI deployment on cloud marketplaces.

What’s Included

• Pre-installed NVIDIA GPU drivers and Container Toolkit

• Optimized for standalone NVIDIA AI containers

• Available on AWS, Azure, Google Cloud, and Oracle Cloud Infrastructure

• Supports all standard GPU instance types listed in Standard GPU Instances

Table 17 NVIDIA GPU Optimized VMI on CSP Marketplace

Cloud Service Provider	VMI Name	GPUs	K8s Support	Standalone Container
Amazon Web Services	NVIDIA AI Enterprise	Refer to Standard GPU Instances	N/A	Supported
Microsoft Azure	NVIDIA AI Enterprise	Refer to Standard GPU Instances	N/A	Supported
Google Cloud Platform	NVIDIA AI Enterprise	Refer to Standard GPU Instances	N/A	Supported
Oracle Cloud Infrastructure	NVIDIA AI Enterprise	Refer to Standard GPU Instances	N/A	Supported

CPU-Only Server Support

NVIDIA AI Enterprise supports the following CPU-enabled frameworks:

• TensorFlow

• PyTorch

• Triton Inference Server with FIL backend

• NVIDIA RAPIDS with XGBoost and Dask

The CPU-enabled frameworks are supported on CPU-only servers that are part of the NVIDIA-Certified Systems list.

Footnotes

[1] (1,2)

On NVIDIA HGX platforms, NVIDIA supports Fabric Manager for PCIe Passthrough (1-, 2-, 4-, or 8-GPU VMs), shared NVSwitch configurations, and vGPU deployment modes. For implementation details and integration with your virtualization environment, consult the documentation from your hypervisor vendor.

[2]

With optional RTX 6000 Ada Lovelace GPU.

[3] (1,2)

The NVIDIA vGPU for Compute Guest Driver is optional in bare metal deployments. Its use allows licensing enforcement.

[4]

Red Hat OpenShift Virtualization is supported with Red Hat OpenShift versions mentioned in the table and only with GPU Passthrough and NVIDIA vGPU for Compute.

[5] (1,2,3,4,5)

Grace Hopper GH200 and Grace Blackwell GB200 and GB300 support is limited to bare metal deployments.

[6] (1,2)

NVIDIA vGPU for Compute provides advanced management and monitoring capabilities that are not available with GPU Passthrough and the NVIDIA Data Center GPU Driver.

[7] (1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18)

NVIDIA vGPU for Compute software is supported on Linux with KVM platforms only by specific hypervisor software vendors. For information about which NVIDIA vGPU software releases and hypervisor software releases are supported, consult the documentation from your hypervisor vendor. For information about Linux distributions supported as a guest OS, consult the documentation from your hypervisor vendor. NVIDIA vGPU for Compute software and your virtualization software vendor must support a guest OS release.

[8] (1,2,3)

NVIDIA AI Enterprise supports every patch release for the listed Red Hat OpenShift release, provided that Red Hat also supports it. When Red Hat no longer supports a release or patch release, it is no longer supported by NVIDIA AI Enterprise.

[9] (1,2,3,4,5,6,7,8)

Obtain the NVIDIA Virtual GPU Manager software directly from Nutanix through the My Nutanix portal (a My Nutanix account is required). Contact Nutanix if the NVIDIA AI Enterprise release is unavailable from the My Nutanix portal.

[10] (1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17)

NVIDIA AI Enterprise only supports running workloads on a single-node setup for KVM hypervisors.

[11]

Refer to the NVIDIA GPU Operator documentation for the list of Government Ready operands.

[12] (1,2,3,4)

Refer to the NVIDIA Network Operator documentation for the list of Government Ready operands.

[13]

For B200 HGX, vGPU for Compute support is enabled on Linux KVM hypervisors only.

[14] (1,2)

On NVIDIA HGX platforms, vGPU for Compute VMs with configurations of 1, 2, 4, or 8 GPUs per VM is supported only with VMware vSphere.

[15] (1,2)

On NVIDIA HGX platforms, GPU Operator is not supported with KubeVirt or OpenShift Virtualization.