Here is a link I posted on the old forum on how to perform software pwm on daisy chained 595s using SPI.
http://www.arduino.cc/cgi-bin/yabb2/YaBB.pl?num=1278251011
Also, if you don't want to use SPI for some reason, may also check out using direct port manipulation to speed up your shift out.
http://www.arduino.cc/cgi-bin/yabb2/YaBB.pl?num=1254488619
Good luck and have fun.