AWS Marketplace AMI

written by Darren Lauser, edited by Alan Cugler on 2023-10-01

Terminal Labs Blog Posts:

When using an AWS instance with a physical NVIDIA GPU, the driver is not ready to use out of the box. AWS provides instructions, and the install package, for getting the driver installed and optimized, but it is not a trivial process.

Terminal Labs has used AWS for a decade and grown accustomed to EC2 provisioning both from the AWS console and within the VM. However, we realize there are lots of people that don't have the time or skill to setup these VMs for their work in data science, engineering, and Machine Learning, among other fields. So to simplify some of the provisioning we have started building AMIs for people to use in the EC2 provisioning process with prebuilt GPU drivers installed for the applicable NVIDIA GPU offering AWS has listed.

We also intend to continue to specialize AMI offerings for specific use cases so practitioners can have a targeted solution for their needs. Contact us for developing your specific AMI needs!

If you need an AWS GPU instance with Ubuntu 22 and the NVIDIA GRID driver already installed and optimized, please consider subscribing to our AWS Marketplace AMI.

Sometimes it takes AWS some minutes (up to an hour, they say) to process the subscription request, but this only needs to be done once. So, you may want to subscribe now, to get that out of the way.

One other step that could take some time, and this one could take hours for AWS to process, is to raise your limit for the number of g4dn instances you are allowed to launch. For some accounts, the default limit is zero, so you will not be able to launch this instance type until after the limit has been increased. This limit applies to each region in each account. Service Quotas -> AWS services -> Amazon Elastic Compute Cloud (Amazon EC2) -> search for g4dn.

Disclaimer:

NVIDIA, and the NVIDIA logo, are registered trademarks of NVIDIA.