There are no analogue stages to consider and you have total control of the dynamics at source so no, it doesn't matter.
(This is assuming you are streaming the soft synth audio via a digital connection, which you are implying, I think!).
A lot of people have experienced a favourable difference when running softsynths at a higher sample rate though, so you should try having the source machine set to 96k for example, and record that audio in.
If you don't want your recording machine set at 96k, then you'll have to convert the sample rate down using something like Voxengo R8 Brain.
But, you'll have to render the v.i., not record in real time.
Otherwise, run the softsynths at the higher sample rate and send out via analogue, perhaps through some sort of cheeky line-driver (I use a vtb-1 but only have one of them - drag) and then in to your 'tape machine' rig, which will take you back to observing sensible and appropriate input levels again
What's the rationale behind feeding digital sources in this way, is it a slave machine running cpu hungry plugs? (Just curious).