Here is what I remember doing back when I was using a parallel cable.
There were 3 bios settings for parallel, and the "fastest" one involved DMA. I always found that setting to grab too much processor time and otherwise stall the computer. I used the "slower" settings and they allowed the other programs to continue with little or no change to print time.
When the spooler settings allowed, I found that "finish spooling then print" was much more effective than otherwise. This sometimes had a print take longer to start, but acually completed faster.
Generally speaking, it seems that the printer manufacturers have stopped developing thier parallel drivers around the time of windows 95B, so they have lost efficiency in any newer OS.
I believe the Canon Prixima 4000 is about $80 now, usb, fast print, cheap ink ... Just a thought.
Jim