There are two online servers available, I-TASSER (http://zhanglab.ccmb.med.umich.edu/I-TASSER/) and Robetta (http://robetta.bakerlab.org/). Offline you can use Rosetta software. 

Consider also looking at

https://www.predictprotein.org/

There are things you should do before ab initio structure prediction. Is the protein disordered? (You can try DisoPred or simply calculate the % charged residues and compare to similar sized globular proteins.) Does it have transmembrane regions? (Try TMHMM) Signal sequences or motif patterns? (Prosite) Sometimes you can find a significant Blast hit using the Prosite pattern (PHI-Blast). Finally, check the literature for information about not just your protein but all of its homologs in PFAM. Maybe somebody knows something that will be useful to you. AB initio will give you a structure, but not a reliable confidence metric. If I read you correctly, the sequence has a homolog of known structure. In that case, you need homology modeling, not ab initio.