By block 400k or so the node becomes network bound. No addition of threads speeds (or slows) the sync beyond this point. However less than 8 threads (tested at 4 and 2) is insufficient to reach this limit. In that case speed is reduced by a constant about (i.e. scaling per thread).