General rule: keep your data separated from your application. There are many reasons:
- You can have multiple instances of your application working from the same data set
- Upgrading your application will not affect your data
- CPU is dedicated to one thing: data for data, application for application, choose appropriate sizes for each
- Having read-only replicas for performance boosts
Once you've done that, then it's a question of:
- EC2 Application server + RDS SQL Server, or
- EC2 Application server + EC2 SQL Server
That answer to that purely comes down to management and whether you want to be responsible for that or not (applying patches, setting up read-only replicas, setting up multi-AZ, etc.)
RDS allows you to easily set these up. But you're limited to what RDS provides. Running your own SQL Server gives you more flexibility, but you have to manage everything.