In our department, we don't have one big prelim exam that tests our graduate students on what they've learned in their first year or two of grad school. We space it all out over the course of their first year and a half in "cume" exams, on which they need to get a certain number of points in a certain number of tries in order to pass. They also have to do an original proposal, grant application style with an oral defense. I like this way, the stress is more balanced throughout the year rather than all pinned on one day, and it fits a broader range of learning and performance styles.
I have to give my first cume this fall, and I'm totally scared. I have to come up with a current, compelling set of questions on a current research article in a certain topic area that will balance the rigor of what we expect grad students to be able to figure out against what they are capable of handling. I have a few past exams to look at, but they're pretty different from the kind of thing I was planning to do--mostly because I'm forging a slightly different research focus path than most of my colleagues have done, and also, I haven't gotten involved in teaching any of the graduate courses yet so I don't have the same frame of reference for what they are expected to know. There aren't any strict guidelines on what and how you have to write these, so I will be exercising some creative license and hopefully won't bomb it.
As with many of these first-timer responsibilities, I feel like it's more of a test for me than for them!