On the basis of several years of dealing with non-technical users and their reactions with OpenID and other single signon technologies, my opinion is that in aggregate non-technical users are very confused by them and confusion leads to failure to convert. The most common user behavior is to create a username/password at your site and use the same credentials they always use. This is what they've been doing for over a decade now. The newfangled options have not been explained to them and they do not understand the fundamental "Facebook Connect replaces having an account with this service" value proposition. For that matter, many of them have a severely broken mental model of what having an account means in the first place. (I have some users who are convinced they need one bingo account for the blue Googles and one for the green Googles.)
I would certainly implement accounts for your service, and test that against Google. This assumes you have 100+ visits a day, which you'd want for A/B testing. If not, I'd implement accounts and deemphasize the Google option (e.g. give it a simple text link when the signup form was prominently visible with big colorful buttons). You can always test later after you have some traction.
With regards to technical users, I have no data which I can share with you but I strongly, strongly suggest rethinking any strategy which does not have internal accounts unless you are committed to customer acquisition via spam-early-spam-often viral mechanics.