The overwhelming consensus is that obviously running in dual channel is faster, but it's open for debate just how much faster. Honestly, I think only when running programs that are really memory intensive on two systems side by side (one in single, the other in dual) would you really see a difference. Benchmarks have been done and users say their applications seem faster but keep in mind that's with them knowing they're running in dual vs. single (blind test anyone?).
usually from tests 1-5%, but it depends on the hardware used. If the system is a Llano or Sandybridge that is using the IGP from graphics output then the system is likely to receive a larger boost otherwise it seems that the boost is negligible and having sufficient memory is more important than being in dual/triple/quad channel mode.