I am seeking advice on relevant solution/database for extracting all PRDM9 motifs across human genome. Ideally I'd like the coordinates of start position and end position for such motifs across all human chromosomes.
You might look for a PRDM9 motif model matrix from one of these TF databases: http://floresta.eead.csic.es/footprintdb/?databases
Once you have the model's matrix in MEME format, you could run that through FIMO to find putative binding sites (in BED format): http://meme-suite.org/doc/fimo.html
EMBOSS fuzznuc might solve your issue. The output isn't exactly in bed format, but it does return start/end coordinates in a table-formatted file and also allows for fuzzy searching, which is probably handy considering the degeneracy of the PRDM9 binding motif. (Caveat being that I haven't used it myself, however)
Edit: apologies, I realize your question is also asking about a database from which these motifs could be pulled. If I'm understanding that correctly, would the fasta sequence of the human reference genome work?