If I had to hazard a wild guess, I would say that it is because much of  Q3's structure is broken up into nice separate phases, like game, cgame,  front end, back end.  The P4 relies on a trace cache for best performance,  and it would have effects in some ways like a smaller instruction cache. It may turn out to be something unrelated, like a better bus interface helping the graphics driver more.

John Carmack

Speaking of the P4, I guess its about time I start getting ready to re-vamp the Shack benchmarks page. Gotta sucker someone at Intel to send me a P4 though first.