This thread covered this topic back in March:
What Difference Does Each Type Of Winglet Make? (by RootsAir Mar 18 2005 in Tech Ops)
To reiterate my summary:
For a first-order analysis of the benefit from a winglet, lay it down flat as if it was a span extension. Hence, taller is better. Then, look at whether the winglet joins the wing in a sharp break or a smooth flow. Smooth is better.
Using this criteria, I'll rank the winglets (and raked tip):
1) The A3100/310/318/319/320/321/380 "winglets" are the least effective, but also increase wing bending moments the least. I have never seen a published number for how much they reduce drag (even Jupp's patent US4,714,215 says little), but I'd guess they are worth 1-1.5% drag improvement.
2) The MD
-11 winglets come next. They are a direct outgrowth of Whitcomb's original designs. I have some NASA reports on the flight testing of a similar winglet on a DC-10. According to CR-3704, these winglets are produce a 2-2.5% drag reduction in cruise.
3) The Valsan/Quietwing 727 winglets are a latter Whitcomb influenced design. Coupled with a flap droop on the 727, they are reputed to be worth about 5%. The flap droop reduces wing bending moments and helps reduce wave drag, so the 5% is a mixture of effects. The winglet by itself is probably worth about 3%.
4) The 747-400 and A330/340 winglets are rather similar. They fall on the wing span extension line. I've read that the 747-400 winglets produce about a 3.5% drag reduction.
5) The 776-400ER/777-300ER/777-200LR raked tips also fall on the span extension line. They produce varying amounts of drag reduction, depending on their span. They also increase wing bending moments quite a bit.
6) The API/APB blended winglets produce better results than the span extension line indicates. The 737NG winglets are first generation and I've read that they produce about a 4.5% drag reduction. Later API/APB winglets improve on this.