docs(aws): added info on p2 instances, refactored to own section

docs/aws.md

@@ -50,22 +50,24 @@ The goal of this document is to provide a guide on how to set up the GitHub Acti
       - `GITHUB_TOKEN` - The GitHub token you copied earlier.
 4. Choose an (or create) an AMI
     - We recommend Ubuntu 22.04 to stay in-line with [GitHub Actions](https://github.com/actions/runner-images#available-images)
-        - For P2 instance types we have created an AMI to work with the CUDA driver for the Nvidia K80, see the table below.
-        - For all other Nvidia GPU instances, we recommend using the [Amazon Deep Learning AMI](https://aws.amazon.com/blogs/machine-learning/get-started-with-deep-learning-using-the-aws-deep-learning-ami/)
     - To ensure compatibility, ensure that `docker` and `git` are installed on this machine
     - To create your own AMI please review these [AWS docs](https://docs.aws.amazon.com/toolkit-for-visual-studio/latest/user-guide/tkv-create-ami-from-instance.html)
+    - Please see below for more information on recommendations for GPU instances
 
 **NOTE**: If you are already using AWS for EC2, you may consider creating a [VPC](https://docs.aws.amazon.com/vpc/latest/userguide/create-vpc.html), [subnet](https://docs.aws.amazon.com/vpc/latest/userguide/create-subnets.html), and a [security group](https://docs.aws.amazon.com/vpc/latest/userguide/working-with-security-groups.html) with outbound traffic on port 443 to isolate your runners from the rest of your AWS account.
 
 You are now ready to start using this action with AWS!
 
-### OMSF p2.xlarge AMI Table
+### GPU Instance Recommendations
+- If you are using almost anything other than the P2 instance type, we recommend the use of [Amazon Deep Learning AMI](https://aws.amazon.com/blogs/machine-learning/get-started-with-deep-learning-using-the-aws-deep-learning-ami/)
+    - For compatibility and recommendation from AWS, see [here](https://docs.aws.amazon.com/dlami/latest/devguide/gpu.html)
+- The cheapest GPU option available (from P instance types) is the `p2.xlarge` however, these instances do not have support from AWS for the new Deep Learning AMI. We have created and hosted an AMI with the proper CUDA installed with Docker pre-bundled. See below for those AMIs.
+- **P2 instances use version 470 for the Nvidia kernel driver. As a result, we only ship CUDA v11.4 in the AMI(s) above. You _must_ pin your `cudatoolkit` to be 11.4 to use this instance type.**
+#### OMSF p2.xlarge AMI Table
 | AMI | Region | Instance Type |
 |-----|--------|---------------|
 | ami-073f98140576b5a81 | us-east-1 | p2 |
 
-**NOTE**: P2 instances use version 470 for the Nvidia kernel driver. As a result, we only ship CUDA v11.4 in the AMI(s) above. You _must_ pin your `cudatoolkit` to be 11.4 to use this instance type.
-
 
 
 ## Additional notes for requesting GPU instances on new accounts