Sorry, you're wrong on this. If you've got one CPU, then running one instance of lame at a time will go faster than running two together. The reason for this is that lame is CPU-bound. It pegs the CPU meter. If you run two copies of lame together, each will take almost precisely twice as long to finish.

The best way to speed things up is to use EAC on Windows or grip on Linux, which run lame in the background and create a work queue of tasks for lame to process. This way, my blazing Plextor drive can rip audio accurately at 16-20x, allowing me to plough through 20 CDs in an hour. Then, the computer grinds away running lame all day and I've got fresh MP3s waiting for me when I get home at night.