Fold recognition by combining sequence profiles derived from evolution and from depth-dependent structural alignment of fragments.

Try
PHYRE Protein Homology/analogY Recognition Engine Protein Fold Recognition Server

DescFold(Descriptor-based Fold Recognition System)

or Pcons-meta server

Consensus Fold Recognition Methods [PPT]

Cascade PSI-BLAST web server: a remote homology search tool for relating protein domains

Also I remember a old Perl program called "Divide-and-BLAST"

As suggested in the first answer, it is always good to throw a range of fold prediction methods at your sequence. Some may perform better than others and some confidence is gained if several methods give comparable results.

However, for an uncharacterised protein, often prediction can only take you so far. As you mentioned, the OB fold is common to many proteins: here's the SCOP fold entry and the SSB family entry. Fold/domain information can be useful but sometimes, you just can't go any further with computational analysis and it's time for the experimentalists to up their game and generate more data!

Phyre has a great web interface with many options to visualize data.

If you want to try some other methods for protein fold prediction, visit the web site of the CASP competition. You will find many useful tools (perhaps too much).