Question title says it all. It seems to me like the only difference is that HPC Optimized instance types are a bit more more optimized to access storage. What's very confusing is that the Compute Optimized instance type has high performance computing in its list of use cases.
Can anyone explain the real difference between the AWS Compute Optimized and HPC Optimized EC2 Instance types?
You can use the compute optimized instances (eg C6, C7 etc) for HPC. It's just that the HPC instances are (a) better optimized for those workloads (we do some neat stuff inside those Hpc7x/Hpc6x boxes); (b) they have more plumbing back to parallel storage; (c) they exist in larger capacity lumps that're well connected to each other to make EFA work even better.
Also, Hpc6/7 instances only exist in single AZ in each AWS region they're provided in. We do that to consolidate the capacity so we don't end up with fragmentation issues, and HPC customers by and large tell us that's ok, because they're not looking for redundancy for HPC workloads - they want speed.
Oh, and they're also like 65% cheaper than the other instances. Almost forgot that bit :-)
Oh, #iworkforamazon in case that wasn't obvious.