The answer is not simple I'm afraid. I've had a lot of users in remote high latency locations who actually prefer the Ajax client because to them it seems much faster compared to the plugin. I believe some of this is because the IE Plugin has single threaded operations which are waiting for a response from the server, so the UI is being locked / frozen in these high latency environments, whereas the Ajax plugin does not freeze up, so it seems faster. I believe some of this performance disparity may also be affected by the volume of data being physically transferred to the client. This will vary depending on what objects you use in the UI, and your data model.
You also need to consider the functional differences between the clients as well (e.g. some macros will not work in Ajax), and the quality of the layout from a visual perspective will be impacted by the amount of consideration the developer has given to the Ajax client (for a lot of developers it is an afterthought at best).
Regarding the performance speed difference - it will depend on the environment, the design of the QV application, data model, and the actions being performed, so I would suggest doing some tests for your specific scenarios to get representative figures. If someone gives you a percentage difference for performance, the probability of it being relevant or representative of your environment and scenarios is very low.
Probably not what you were looking for, but it's more nuanced than a single figure and is impacted by so many variables, which might help to explain why support gave you a wishy washy answer.