"If I understand correctly - you're essentially arguing that rendering on the client is faster because the server has to generate more content and ship more content over the wire with a HTML version? If this is right then you need to set up the test to optimise both cases, not just the client-side template case. May be all this leads to a result that there really is no difference but proving the client-side is faster while running the other case in a non-optimal config doesn't help e.g. is the extra download really 494ms with gzip on?" - Andy Davies