Is this true "de-dup" i.e. at file segment or block level or is it simply single instance?

What about impact on restore times if pulling back from tape? Performance metrics?

We're not that big of a shop, and our data protection scheme starts with a disk-to-disk backup for easy recover of the near-realtime data, with the disk storage then being run to tape. Since the de-duping is easily done on disk, what's the need for the tape to do it?