My laptop has a Broadcom 4311 rev 01 wireless chipset. The drivers from the latest Linux git releases are vastly superior to the old bcm43xx driver... so as of yesterday, I'm back to running the latest wireless-2.6 code. The former upstream maintainer claims that he gets better throughput with the reverse-engineered Linux driver than he does on Windows XP.
While I was at it, I compiled in dynamic tick (tickless) support. I then had some fun with powertop, and managed to bring CPU wakeups down to about 8 per second. It seemed that using the framebuffer console required 5 wakeups per second, but the non-framebuffer one needed around 250... this was dreadfully unscientific, though.