If you're at all interested in this topic, I can't recommend highly enough a little book by Tony Kordyban, Hot Air Rises and Heat Sinks: Everything You Know Abo

If you're at all interested in this topic, I can't recommend highly enough a little book by Tony Kordyban, Hot Air Rises and Heat Sinks: Everything You Know About Cooling Electronics Is Wrong. (The sequel, More Hot Air, is good too). The author makes the case, in Chapter 30, that those numbers, which came from MIL-HDBK-217, were pulled out of somebody's hat at a time when semiconductor reliability rates were changing rapidly due to process improvements, and probably don't have much validity any more. When actual reliability rates were compared to prediction and measured temperatures, there was no correlation. These books are great. Refs: [1] Tony Kordyban. Hot Air Rises and Heat Sinks. New York: ASME Press, 1998. ISBN 0-7918-0074-1 [2] ---. More Hot Air. New York: The American Society of Mechanical Engineers, 2005. ISBN 0-7918-0223-X