Quicksight connection with Redshift RA3 or DC2, which is better?

For QuickSight dashboards , is it better to have an RA3 cluster or a DC2 cluster ? If someone has used both , can you please comment on the data refresh/load times for both these kinds of clusters ?
I know that DC2 keeps all the data locally , whereas RA3 keeps some data locally as well as some on S3 , will that be a factor in refresh times of the datasets?

Clusters based on ra3 unlock AQUA-based query improvements for string matching, but you have to setup AQUA on the cluster, AND you have to setup your redshift query so that it matches the AQUA wildcard-search query patterns. AQUA creates a cache that speeds up certain queries in the Redshift cluster.

From what I recall, ra3 is generally faster too since it’s a newer class of machine, but I don’t have metrics to support that offhand.

As for which is better: A vs. B, the answer will differ of course per customer, per dataset, and per environment. I encourage experimentation where possible to discover what works best for your use case.