Anecdotally after migrating from RDS MySQL to serverless Aurora there was a noticeable slowdown of our dashboard and reporting tools. Our typical workloads (ecommerce transactions) are slightly slower on average, but the peaks seem down.

Serverless Aurora likely has different performance characteristics than normal Aurora.

I never did a thorough benchmark but I was working with a poorly indexed DB (actually...no indexes) that had millions of records and despite the lack of indices the database still queried quickly.

Read speeds are pretty good but write speeds not so much. For our particular use case 99.99% of queries ran would have been inserts with reports only generated once per month.

What was the concurrency like in those benchmarks?

I don't remember as this was 2 years ago. We were only concerned with write speeds as that was the majority of queries we would be performing. Read speeds were pretty good.