I don't see my reply in this thread - I thought I posted one; well anyway:
Is noise the same across the board?
What I meant to ask is, would the noise levels resulting from using a lower ISO with a longer exposure be the same as that same exposure using a higher ISO and correspondingly shorter exposure?
The important things to realize are:
1) Dark current noise (related to exposure time and temperature) exists relative to signal in the analog sensor wells, and the ratio has absolutely nothing at all to do with the ISO setting.
2) ISO-related noise exists relative to RAW exposure (the dynamic range of the RAW data at that ISO).
3) Depending on the camera, there may be more noise at higher ISOs relative to absolute signal, the same, or less noise.
#3 is very important to know for your camera, but I don't know if this data is actually listed anywhere for specific cameras. I know that for recent Canon models, the lowest ISOs have the highest noise, relative to absolute signal. This means that you can consider absolute exposure first, and use ISO accordingly, to make sure that the ISO chosen lets your signal give a good exposure for that ISO. If your camera does not improve in noise efficiency at higher ISOs, then it may not be worth using the higher ISO, since it will not give you any less noise.
As far as the trade-off between time and signal-to-dark_current_noise is concerned, in most of my experience, dark current noise does not quite double in double the time, so the longer exposure is generally better; IOW, with the same f-stop, you are better off with 8x the exposure time at ISO 200 than 1x at ISO 1600. (ISO 100 is not always a good ISO for highlights, depending on the camera).
A more objective experiment, however, would be in order. I would expect to see any advantage for the shorter exposure (all other things being equal) to only happen at very long exposures (multiple minutes), because there is *always* more noise with a higher ISO or under-exposed/pushed lower ISO that has nothing to do with dark current. In my recent Canon 20D test, ISO 100 at 30 seconds was better than ISO 1600 at 2 seconds, but clearly part of the benefit was the lower non-dark_current noise at ISO 100. The hottest pixels were slightly hotter in the 1600 image.