If the sequences that you are trying to annotated are functional and distant, using blastx would provide better results than just blastn, specially to annotate distant but functional related proteins.

I don't think that the poster ever wanted to use blastn - the question was about translating the sequences before or during the search. My recommendation is to translate the sequences before the search and use blastp, as that will cut down on time.

My apologies, I see that the issue of using blastn was brought up. Still think it is better to translate ahead of time and use blastp.