It does not have to be better at all, like lightning50 said; obviousness and how well something performs in the relative sense are two seperate issues, and something can perform worse and be perfectly patentable, but if something performs well in, say, the marketplace, or satisfies a long unmet need, etc. you may introduce this information into the determination of obviousness. The discussion above appears to be directed toward machines (electrical or mechanical) where utility is nearly always a given, and a long line of cases indicates (to me) that in this area you are only looking for any quantum of utility; however in the area of biotech, your utililty must be specific, substantial, and credible, which is quite different. Before, the idea of using "some quantum of utility" was essentially applied in chemical cases, and people were using this literally, and saying things like a newly discovered or synthesized protein may be used for things like hair shampoo, which is quite ridiculous, so the Federal Circuit changed things around. You can draw novel structures all day that, without some showing of utility that meets these criteria, would not be patentable. And it does not require the use of actual lab results either, although a friend of mine who is an attorney practicing in the biotech area says it does for pharmaceuticals, though my research of the case law makes me think otherwise. Anyway, at least from my point of view, it is best to seperate the utility requirement from obviousness/novelty considerations, and forget about whether something is better or worse in the relative sense.