Max IOPS for non-EBS-optimized EC2 instance [closed]

I am trying to set up a database server using an m3.large EC2 instance, and would like to use a provisioned IOPS EBS backing store to maximize performance. Purchasing an EBS-Optimized instance would be ideal, but the instance size required to get this feature is out of the budget for this project.

Has anyone experimented with or observed a rough max IOPS you can get on a provisioned IOPS non-optimized store? If I set it too high, I am paying for provisioned bandwidth that the machine could never possibly leverage over its connection, and if I set it too low, I am unnecessarily choking my instances access to its data. I was considering provisioning 500IOPS, but I don't have any evidence to back this number other than guessing based off the IOPS count used in examples for EBS-optimized instances.

EBS averages 100 IOPS but can "burst to hundreds" per https://aws.amazon.com/ebs/details/. The ability to burst indicates that the network bandwidth (used by EC2) is potentially there for 500 PIOPS to be used, depending on the other EC2 instances on the same host.

Your latency is going to vary more than with an EBS-Optimized instance. PIOPS on EBS-Optimized instances work by having dedicated network bandwidth straight to the EBS servers. On a non-optimized instance, other network traffic from your instance or your instance's neighbors can affect the EBS traffic.