I agree about the AD8512, ppl, more or less. I don't know if I'd call it grainy, but instead call it hyper-revealing. I don't have a whole lot of listening time on this chip, but everything bad I've heard with it I'm convinced exists in the recording.
These faults might be enough to dismiss the chip, except for a few factors:
1. It runs from lower voltages than most other chips, including the 8610 and the 845, which is important when running from a single 9V battery like a lot of my amps have to do. Although this chip isn't classed as a rail-to-rail chip, output voltage can still come quite close to the rails, especially when the load is high impedance, as in the META42.
2. The low supply current is really quite remarkable. It's lower than the already low AD823.
3. This chip doesn't sound as aggressive to me as the 823.
4. The 845, my current favorite for high-end single-9V amps, sounds muffled compared to the AD8512. To put it another way, the 845 is smooth to the point of lacking detail, and the 8512 is detailed to the point of being unforgiving. Perhaps some day I'll find a middle ground between these two that still runs well from a single 9V battery.
These factors have caught my attention enough that I may stop using AD823s in favor of the 8512. Chips at this level of performance are all imperfect, but the 8512 seems to have a good balance of imperfections vs. good features.