how to select EC2 Instance with NVLINK support

To select an EC2 instance with NVLink support, you should focus on specific accelerated computing instance types that are designed for high-performance machine learning and AI workloads.

Based on the available information, the following EC2 instance types support NVLink:

1. P3 instances (except p3.2xlarge):

• p3.8xlarge: 4 NVIDIA Tesla V100 GPUs with NVLink
• p3.16xlarge: 8 NVIDIA Tesla V100 GPUs with NVLink
• p3dn.24xlarge: 8 NVIDIA Tesla V100 GPUs with NVLink

2. P6e-GB200 UltraServers:

• These instances are accelerated by NVIDIA GB200 NVL72
• Feature 1800 GB/s peer-to-peer GPU communication with NVIDIA NVSwitch
• Offer the highest GPU AI training and inference performance in Amazon EC2

When selecting an instance with NVLink, consider your specific requirements for:

• Number of GPUs needed
• GPU memory requirements
• vCPU and system memory needs
• Network bandwidth requirements
• Budget constraints (On-Demand vs. Reserved Instances)

To find EC2 instances with NVLink support, you can:

1. Review the accelerated computing instance families in the EC2 console
2. Use the EC2 instance type finder to filter for GPU-accelerated instances
3. Check the detailed specifications for each instance type to confirm NVLink support

NVLink is particularly valuable for workloads that require high-bandwidth communication between GPUs, such as distributed deep learning training and high-performance computing applications.
Sources
Cloud Compute Instances – Amazon EC2 Instance Types – AWS
Amazon EC2 P3 – Ideal for Machine Learning and HPC - AWS

Hi,

To find EC2 instances with NVLink support, the most efficient method is to use the AWS CLI with the describe-instance-types command, filtering by instance type and then using grep to find instances with "nvlink" in their GPU information. Alternatively, you can manually review instance details in the AWS Management Console, though this is less efficient.

Step 1: Install and Configure AWS CLI Install and configure the AWS CLI on your system with appropriate AWS credentials. This is a prerequisite for using the CLI method.

Step 2: Execute the AWS CLI Command Run the following command in your terminal:

aws ec2 describe-instance-types --filters "Name=instance-type,Values=p4d.24xlarge,p5.48xlarge,g5g.16xlarge,g5.48xlarge,g4dn.metal,p3dn.24xlarge,p3.16xlarge" --query "InstanceTypes[*].{InstanceType: InstanceType, GPUInfo: GpuInfo}" --output text | grep -i nvlink This command filters instance types, extracts GPU information, and searches for "nvlink" in the output. Adjust the Values in the --filters argument to include instance types relevant to your needs.

Step 3: Interpret the Results Examine the output of the command. Instance types listed in the output that contain "NVLinkInfo" sections support NVLink. The NVLinkInfo section provides details about the NVLink configuration (e.g., number of lanes).

Step 4: (Alternative) AWS Management Console Go to the EC2 service in the AWS Management Console. Click on "Instances" then "Launch Instance". Select an appropriate AMI. In the "Choose an Instance Type" step, filter by "GPU instance". Manually review the details of each GPU instance type to check for NVLink support in the description or specifications.

If the Answer is helpful, please click Accept Answer and up-vote, so that it can help others in the community looking for help on similar topics.