That article is a joke, totally unscientific, his testing methods don't isolate cpu-bound tasks from i/o-bound tasks and his graphs don't even indicate what units he's measuring, probably because there is no unit for subjective user perception.
I'll run some i/o tests using real testing tools once I get it built, but there's no way a single disk will come close to matching read and write throughput compared to a raid-0, all other things being equal.
Now, it's possible my video editing tasks are all heavily cpu-bound, in which case I won't get much benefit from faster i/o for those tasks, but I'm not sure about that yet.
Btw, I've been building raid arrays since 1998 =)
Super Cheap - Canon ZR900, Aiptek A-HD, Premiere Pro CS4, Avidemux, VirtualDub, Audacity, GIMP