On the Alpha 900, ISO 320 is a sweet spot and ISO 160 is the optimum setting. I won't revisit to quote the charts, but DxO's tests showed that the settings below 200 on the A900 are non-linear; there's not really much difference between 100, 125 and 160. The same broadly applies to the A77 settings from 50 to 80, the sensor is being overexposed and the gain adusted to compensate.
For some reason, the A to D gain seems to be adjusted in full steps only - 200, 400, 800, 1600 - on the A700/900/850. The intermediate steps are achieved by changing the digital gain which comes after A to D stage. From 2000 to 6400, all gain is digital. I don't know exactly why this happens, but it seems that 320 is achieved by reducing gain from 400, 160 by reducing gain from 200 (I use 640 as well for the same reason). The effect is to improve the noise more than you would expect, so much that 320 can look smoother than 200. Certainly there's no real point in using 200 as a setting, 320 generally looks just as good, and 160 is as good as 100 for noise but with slightly less risk of clipping the highlights - better recovery from raw, maximum dynamic range.
What really gets me about the A900 is the difference in colours. I quite like the A77, but there's a quality to the saturated reds and greens which is artificial looking compared to the A900. Canon's colours in the 5D Mk III are not pleasant at all (the earlier 5D models were both better).
Generally, I don't like Nikon colours as much as Sony. That's the dilemma. New cameras can give me higher ISOs, greater resolution, but no matter what colour profiles I apply and what processing settings I use, I can not get the same effortlessly fine accurate colour rendering the A900 provides.
I suspect this is because the A900 uses denser RGB filtration. Actually, I know this is the case. That is what my very first comparison test of the Nikon D3X and the A900 showed. All the extra high ISO noise of the A900 was caused by two things - first of all, the high ISO was a truer setting, the Nikon typically giving 1/3 to 2/3rds more actual exposure to get the same brightness values; secondly, the chroma content of the Nikon files was reduced.
Since then, every maker including Sony has taken the route of using lower density colour filters on the sensor, producing less discrimination but paradoxically lower Delta-E values (colour error - with less colour to start with, colour errors are reduced in their deviation). Fuji has even doubled the number of green sensors in the X-Pro1 etc chip. Colour is constructed in software. It's even reduced in intensity as the ISO goes above A-to-D gain levels, normally over 1600 or 3200, in all camera models - low light shots have a monochromatic look. Sony has resisted this trend as much as possible and even with changes, keeps pretty good colour across the normal range of ISO equivalents.
But... the Alpha 900/850 and 700, the 100/200/300 series with CCDs, the Sony DSC-R1 - all of these use full strength RGB on the sensor, just like the KM7D and 5D did before. The first model with a visibly different colour filter array was the Alpha 350, and for that camera, they didn't try to boost the colour artificially. It has a naturally more pastel rendering which is very subtle, like a portrait-type neg film. They also didn't try to optimise its high ISO.