NVIDIA AI Enterprise Infra 6.0 Support Matrix

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

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 in the following content

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

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

Requirements

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

On-Premises Servers

1. GPU supported by NVIDIA AI Enterprise (refer to Supported GPUs table in the following section)

2. NVIDIA-Certified system

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

Cloud Servers

1. Supported orchestration platform (refer to Public Cloud section in the following content)

2. Supported virtual machine instance - standard GPU instances

Supported NVIDIA Infrastructure Software

Table 93 Supported Infrastructure Software

Product	Version	x86	ARM
Core Infrastructure Drivers
NVIDIA GPU Data Center Driver	570.86.15	Supported	Supported
NVIDIA DOCA-OFED Driver for Networking	2.9.1	Supported	Supported
Container Platform
NVIDIA Container Toolkit	1.17.4	Supported	Supported
Kubernetes Operators
NVIDIA GPU Operator	24.9.2	Supported	Supported
NVIDIA Network Operator	24.10.1	Supported	Supported
NVIDIA NIM Operator	1.0.1	Supported	Supported
Cluster Management & Orchestration
NVIDIA Base Command Manager (BCM)	10.24.11	Supported	Supported

Note

NVIDIA vGPU for Compute (Virtual GPU Manager and Guest Drivers) isn’t supported with NVIDIA AI Enterprise 6.0. A future NVIDIA AI Enterprise 6 minor release will add support for NVIDIA vGPU (C-Series).

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 OS or GPU; refer to the individual product release notes for any discrepancies.

NVIDIA Hopper

• NVIDIA H800

• NVIDIA H800 NVL

• NVIDIA H200 NVL [1]

• NVIDIA H100

• NVIDIA H100 NVL [1]

• NVIDIA H20

NVIDIA Grace Hopper

• NVIDIA GH200

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 Ada Lovelace

• NVIDIA L40S

• NVIDIA L40

• NVIDIA L20

• NVIDIA L4

• NVIDIA L2

• NVIDIA RTX 6000 Ada Generation

• NVIDIA RTX 4000 SFF Ada Generation

NVIDIA Turing

• NVIDIA T4

• NVIDIA T4G

• NVIDIA Quadro RTX 8000

• NVIDIA Quadro RTX 6000

• NVIDIA Quadro RTX 4000

NVIDIA Volta

• NNVIDIA V100

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

Table 94 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 BlueField-3 SuperNIC	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, which is included in the DGX OS software.

Table 95 Supported Platforms

Accelerated Platform	Architecture
NVIDIA DGX B200	NVIDIA Blackwell
NVIDIA DGX H100	NVIDIA Hopper
NVIDIA DGX A100	NVIDIA Ampere
NVIDIA HGX B200	NVIDIA Blackwell
NVIDIA HGX H800	NVIDIA Hopper
NVIDIA HGX H200 [1]	NVIDIA Hopper
NVIDIA HGX H100	NVIDIA Hopper
NVIDIA HGX H20	NVIDIA Hopper
NVIDIA HGX A800	NVIDIA Ampere
NVIDIA HGX A100	NVIDIA Ampere
NVIDIA IGX Orin [2]	NVIDIA Ada Lovelace

Note

DGX platforms, HGX platforms with KVM hypervisors, and IGX Orin aren’t supported with NVIDIA vGPU (C-Series).

For NVIDIA IGX, refer to the NVIDIA AI Enterprise - IGX Packaging, Pricing, and Licensing Guide.

Bare Metal

Refer to the following platform support matrix for NVIDIA AI Enterprise if you have a dedicated physical server on-premises.

Table 96 Containers Orchestrated with Kubernetes on Bare Metal

Orchestration Platform	Versions	Engine	Operating System	OS Versions	GPU Operator	Network Operator	**GPU Driver Support** [3]
Charmed Kubernetes	1.28 - 1.30	Containerd	Ubuntu	20.04 LTS, 22.04 LTS	Supported	Supported	Data Center
HPE Ezmeral Runtime Enterprise	5.6	Containerd	Red Hat Enterprise Linux	8.6, 8.8, 8.10	Supported	N/A	Data Center
Red Hat OpenShift [4]	4.13 - 4.16	CRI-O	Red Hat CoreOS	4.13 - 4.16	Supported	Supported	Data Center
Nutanix NKP	2.12 - 2.13	Containerd	Ubuntu	20.04 LTS, 22.04 LTS	Supported	N/A	Data Center
Upstream Kubernetes	1.24 - 1.30	Containerd	Red Hat Enterprise Linux	8.6, 8.8, 8.10	Supported	Supported	Data Center
Upstream Kubernetes	1.24 - 1.30	Containerd	Ubuntu	20.04 LTS, 22.04 LTS	Supported	Supported	Data Center
Upstream Kubernetes	1.24 - 1.30	CRI-O	Red Hat Enterprise Linux	8.6, 8.8, 8.10	Supported	Supported	Data Center
Upstream Kubernetes	1.24 - 1.30	CRI-O	Ubuntu	20.04 LTS, 22.04 LTS	Supported	Supported	Data Center

Table 97 Standalone Containers on Bare Metal

Container Name	Engine	Operating System	OS Versions	GPU Operator	Network Operator	**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	N/A	N/A	Data Center
Non-Kubernetes (standalone containers)	Docker/Podman	SUSE Linux Enterprise Server	15 SP2 and later	N/A	N/A	Data Center
Non-Kubernetes (standalone containers)	Docker/Podman	Ubuntu	20.04 LTS, 22.04 LTS	N/A	N/A	Data Center

Virtualized

If a physical server is separated into multiple virtual servers on-premises, refer to the following platform support matrix for NVIDIA AI Enterprise.

Table 98 Containers Orchestrated with Kubernetes on Virtualized Environments

Orchestration Platform	Versions	Engine	Guest OS	Guest OS Versions	Hypervisor	Hypervisor Versions	GPU Operator	Network Operator	GPU Driver Support (vGPU)	GPU Driver Support (Passthrough) [5]
Charmed Kubernetes	1.28 - 1.30	Containerd	Ubuntu	20.04 LTS, 22.04 LTS	Ubuntu [6] [9]	20.04 LTS, 22.04 LTS	Supported	Supported	N/A	Data Center
Charmed Kubernetes	1.28 - 1.30	Containerd	Ubuntu	20.04 LTS, 22.04 LTS	VMware vSphere	7.0 update 2 and later, 8.0 and later	Supported	Supported	N/A	Data Center
Red Hat OpenShift [7]	4.13 - 4.16	CRI-O	Red Hat CoreOS	4.13 - 4.16	Red Hat Enterprise Linux [6] [9]	7.9, 8.6, 8.8, 8.10	Supported	Supported	N/A	Data Center
Red Hat OpenShift [7]	4.13 - 4.16	CRI-O	Red Hat CoreOS	4.13 - 4.16	VMware vSphere	7.0 update 2 and later, 8.0 and later	Supported	Supported	N/A	Data Center
VMware vSphere with Tanzu	TKR 1.23 - TKR 1.30	Containerd	Ubuntu	20.04 LTS, 22.04 LTS	VMware vSphere	7.0 update 2 and later, 8.0 and later	Supported	N/A	N/A	Data Center
Nutanix NKP	2.12 - 2.13	Containerd	Ubuntu	20.04 LTS, 22.04 LTS	Nutanix AOS/AHV [6] [9]	6.5 - 6.10	Supported	N/A	N/A	Data Center
Upstream Kubernetes	1.24 - 1.30	Containerd	Red Hat Enterprise Linux	8.6, 8.8, 8.10	Red Hat Enterprise Linux [6] [9]	7.9, 8.6, 8.8, 8.10, 9.2, 9.4	Supported	Supported	N/A	Data Center
Upstream Kubernetes	1.24 - 1.30	Containerd	Red Hat Enterprise Linux	8.6, 8.8, 8.10	VMware vSphere	7.0 update 2 and later, 8.0 and later	Supported	Supported	N/A	Data Center
Upstream Kubernetes	1.24 - 1.30	Containerd	Red Hat Enterprise Linux	8.6, 8.8	Nutanix AOS/AHV [6] [9]	6.5 - 6.10	Supported	Supported	N/A	Data Center
Upstream Kubernetes	1.24 - 1.30	Containerd	Ubuntu	20.04 LTS, 22.04 LTS	Ubuntu [6] [9]	20.04 LTS, 22.04 LTS	Supported	Supported	N/A	Data Center
Upstream Kubernetes	1.24 - 1.30	Containerd	Ubuntu	20.04 LTS, 22.04 LTS	VMware vSphere	7.0 update 2 and later, 8.0 and later	Supported	Supported	N/A	Data Center
Upstream Kubernetes	1.24 - 1.30	Containerd	Ubuntu	20.04 LTS, 22.04 LTS	Nutanix AOS/AHV [6] [9]	6.5 - 6.10	Supported	Supported	N/A	Data Center
Upstream Kubernetes	1.24 - 1.30	CRI-O	Red Hat Enterprise Linux	8.6, 8.8, 8.10	Red Hat Enterprise Linux [6] [9]	7.9, 8.6, 8.8, 8.10, 9.2, 9.4	Supported	Supported	N/A	Data Center
Upstream Kubernetes	1.24 - 1.30	CRI-O	Red Hat Enterprise Linux	8.6, 8.8, 8.10	VMware vSphere	7.0 update 2 and later, 8.0 and later	Supported	Supported	N/A	Data Center
Upstream Kubernetes	1.24 - 1.30	CRI-O	Red Hat Enterprise Linux	8.6, 8.8, 8.10	Nutanix AOS/AHV [6] [9]	6.5 - 6.10	Supported	Supported	N/A	Data Center
Upstream Kubernetes	1.24 - 1.30	CRI-O	Ubuntu	20.04 LTS, 22.04 LTS	Ubuntu [6] [9]	20.04 LTS, 22.04 LTS	Supported	N/A	N/A	Data Center
Upstream Kubernetes	1.24 - 1.30	CRI-O	Ubuntu	20.04 LTS, 22.04 LTS	VMware vSphere	7.0 update 2 and later, 8.0 and later	Supported	N/A	N/A	Data Center
Upstream Kubernetes	1.24 - 1.30	CRI-O	Ubuntu	20.04 LTS, 22.04 LTS	Nutanix AOS/AHV [6] [9]	6.5 - 6.10	Supported	N/A	N/A	Data Center

Table 99 Standalone Containers on Virtualized Environments

Container Name	Engine	Guest OS	Guest OS Versions	Hypervisor	Hypervisor Versions	GPU Operator	Network Operator	GPU Driver Support (vGPU)	GPU Driver Support (Passthrough) [5]
Non-Kubernetes (standalone containers)	Docker/Podman	Red Hat Enterprise Linux	7.9, 8.6, 8.8, 8.10, 9.2, 9.4	Red Hat Enterprise Linux [6]	7.9, 8.6, 8.8, 8.9, 9.2, 9.4	N/A	N/A	N/A	Data Center
Non-Kubernetes (standalone containers)	Docker/Podman	Red Hat Enterprise Linux	7.9, 8.6, 8.8, 8.10, 9.2, 9.4	VMware vSphere	7.0 update 2 and later, 8.0 and later	N/A	N/A	N/A	Data Center
Non-Kubernetes (standalone containers)	Docker/Podman	SUSE Linux Enterprise Server	15 SP2 and later	VMWare vSphere	7.0 update 2 and later, 8.0 and later	N/A	N/A	N/A	Data Center
Non-Kubernetes (standalone containers)	Docker/Podman	Ubuntu	20.04 LTS, 22.04 LTS, 24.04 LTS	Ubuntu [6] [9]	20.04 LTS, 22.04 LTS	N/A	N/A	N/A	Data Center
Non-Kubernetes (standalone containers)	Docker/Podman	Ubuntu	20.04 LTS, 22.04 LTS, 24.04 LTS	VMware vSphere	7.0 update 2 and later, 8.0 and later	N/A	N/A	N/A	Data Center
Non-Kubernetes (standalone containers)	Docker/Podman	Ubuntu	20.04 LTS, 22.04 LTS, 24.04 LTS	Nutanix AOS/AHV [6] [8] [9]	6.5 - 6.10	N/A	N/A	N/A	Data Center

Base Command Manager Essentials

Table 100 Base Command Manager

Orchestration Platform	Versions	Engine	Operating System	OS Versions	GPU Operator	Network Operator	**GPU Driver Support** [3]
Upstream Kubernetes	1.27 - 1.30	Containerd	Ubuntu	20.04 LTS, 22.04 LTS	Supported	Supported	Data Center
Upstream Kubernetes	1.27 - 1.30	Containerd	Red Hat Enterprise Linux	8, 9	Supported	Supported	Data Center
Slurm (non-Kubernetes)	23.02, 23.11, 24.05	N/A	Ubuntu	20.04 LTS, 22.04 LTS	N/A	N/A	Data Center
Slurm (non-Kubernetes)	23.02, 23.11, 24.05	N/A	Red Hat Enterprise Linux	8, 9	N/A	N/A	Data Center
PBS Pro	2024.1.1, 2022.1.6	N/A	Ubuntu	20.04 LTS, 22.04 LTS	N/A	N/A	Data Center
PBS Pro	2024.1.1, 2022.1.6	N/A	Red Hat Enterprise Linux	8, 9	N/A	N/A	Data Center

Public Cloud

Managed Kubernetes

If you have a virtual server that runs in a cloud computing environment and is accessible remotely, refer to the following platform support matrix for NVIDIA AI Enterprise.

Table 101 Managed Kubernetes

Cloud Service Provider	Orchestration Platform	K8s Versions	Engine	Operating System	OS Versions	GPU Operator	Network Operator	**GPU Driver Support**
AWS	Amazon Elastic Kubernetes Service (EKS)	1.24 - 1.30	Containerd	Ubuntu	20.04 LTS, 22.04 LTS	Supported	N/A	Data Center
Google	Google Kubernetes Engine (GKE)	1.24 - 1.30	Containerd	Ubuntu	22.04 LTS	Supported	N/A	Data Center
Microsoft	Azure Kubernetes Service (AKS)	1.24 - 1.30	Containerd	Ubuntu	22.04 LTS	Supported	N/A	Data Center
N/A	Red Hat OpenShift (Managed Service)	4.13 - 4.16	CRI-O	Red Hat CoreOS	4.13 - 4.16	Supported	N/A	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 102 Standard Instances for Kubernetes and Standalone Containers

Cloud Service Provider	Virtual Machine (VM) Instance with GPU	Product Family
Alibaba	gn7e	NVIDIA A10
Alibaba	gn7i	NVIDIA A10
Alibaba	gn7s	NVIDIA A30
Alibaba	gn6i	NVIDIA T4
Alibaba	gn6e	NVIDIA V100
Alibaba	gn6v	NVIDIA V100
Alibaba	ecs.ebmgn8v	NVIDIA H20
Alibaba	ecs.gn8v	NVIDIA H20
Alibaba	ecs.ebmgn8is	NVIDIA L20
Alibaba	ecs.gn8is	NVIDIA L20
Amazon Web Services (AWS)	EC2 P3	NVIDIA V100
Amazon Web Services (AWS)	EC2 P4	NVIDIA A100
Amazon Web Services (AWS)	EC2 P5	NVIDIA H100
Amazon Web Services (AWS)	EC2 G4	NVIDIA T4
Amazon Web Services (AWS)	EC2 G5	NVIDIA A10G
Amazon Web Services (AWS)	EC2 G6	NVIDIA L4
Amazon Web Services (AWS)	EC2 G6e	NVIDIA L40S
Azure	NCads_H100_v5-series	NVIDIA H100
Azure	NCCads_H100_v5-series	NVIDIA H100
Azure	NCv3-series	NVIDIA V100
Azure	NCasT4_v3-series	NVIDIA T4
Azure	NC_A100_v4-series	NVIDIA A100
Google Cloud Platform (GCP)	A3 VM	NVIDIA H100
Google Cloud Platform (GCP)	A2 VM	NVIDIA A100
Google Cloud Platform (GCP)	G2 VM	NVIDIA L4
Google Cloud Platform (GCP)	N1 VM	NVIDIA T4, - NVIDIA V100
Oracle Cloud Infrastructure (OCI)	BM.GPU3	NVIDIA V100
Oracle Cloud Infrastructure (OCI)	BM.GPU4, - BM.GPU.A100	NVIDIA A100
Oracle Cloud Infrastructure (OCI)	BM.GPU.A10	NVIDIA A10
Oracle Cloud Infrastructure (OCI)	BM.GPU.H100.8	NVIDIA H100
Oracle Cloud Infrastructure (OCI)	VM.GPU3	NVIDIA V100
Oracle Cloud Infrastructure (OCI)	VM.GPU.A10	NVIDIA A10
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

For ease of use in the cloud, NVIDIA provides compute-optimized and validated base Virtual Machine Instances (VMI) to run standalone NVIDIA AI containers through CSP marketplaces. Each VMI includes key technologies and software from NVIDIA for rapid deployment, management, and scaling of AI workloads in the modern hybrid cloud.

Table 103 NVIDIA GPU Optimized VMI on CSP Marketplace

Cloud Service Provider	VMI Name	GPUs	K8s Support	Standalone Container
AWS	NVIDIA AI Enterprise	Listed in Standard GPU Instances	N/A	Supported
Azure	NVIDIA AI Enterprise	Listed in Standard GPU Instances	N/A	Supported
GCP	NVIDIA AI Enterprise	Listed in Standard GPU Instances	N/A	Supported

CPU-Only Server Support

NVIDIA AI Enterprise will support 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] (1,2,3)

NVIDIA H200 isn’t supported by the NVIDIA vGPU for Compute 17 branch.

[2]

With optional RTX 6000 Ada Lovelace GPU.

[3] (1,2,3)

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)

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

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

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.

[7] (1,2)

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.

[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.

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

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