Look up "Banqiao dam failure" on wikipedia, or google it. 26k dead from flooding alone, more than 140k dead from secondary effects. Severe ecological effects and property damage as well. China's got a bad history when it comes to dams.
Even the most severe estimates for Chernobyl are a fraction as many dead, short and long term combined - the highest figure I've ever seen put forward was grossly inflated (the person posting it treated every additional cancer caused by the radiation as "fatal", see if you can spot the logical error there), and it still fell well short of Banqiao in deaths. Fukushima's repercussions aren't fully known yet (Chernobyl's are known because it's been twenty-five years), but there will be far fewer deaths than Chernobyl caused, even according to the people who think Tepco is downplaying the severity.
Other nuclear accidents have single digit fatalities (SL-1 comes to mind), or no fatalities at all. Three Mile Island was a zero casualty disaster, where nobody was killed or irradiated and the final cost was measured in dollar figures alone.
It isn't that nuke plants are intrinsically safe - they aren't. It's that we're so paranoid about nuclear safety we go out of the way when designing for failure, such that the actual damage done by a meltdown is a fraction of what it would be in a plant with few or no safety systems. If we built hydro dams the way we build nuclear plants they'd be incapable of killing anybody when they fail. But we don't. We don't built anything non-nuclear to nuclear-spec safety levels. Which means both the anti-nuke ninnies and the nuclear fanboys are wrong - the former for inflating the danger by pretending there are no adequate safeties and the later for pretending there are no risks.