message queue performance can depend a lot on how you design the structures of your queues.
you should only really do a comparison when you have a representative schema that models the system you want to use. You wont learn much from trying to saturate queue.
you will also find that the queue client you use will effect performance as well as there a few different styles of client that suit lower or higher powered applications. This means using a tool to directly inject messages that isn't the same as the deployed system will give you different characteristics and may not be representative of the real world usage.
I would recommend a combination of modeling real input with real clients to ensure responsiveness (especially with react style clients) and use tools like gatlin to reach the prove the limits of the system e.g. throughput and concurrency.
Using both methods will give you a better understanding of how your application scales and how it feels for users.