You can see a little benchmarking of a 90nm Hammer here.
Essentially, it does a little better than a 130nm chip, and those with it (so far) has gotten this up to about 2.5GHz with high-end air.
It’s a thread worth watching.
This may be stretching the data a bit too much, but engineering samples usually represent the best that can be done at that point in time. ESs usually come unlocked, and those who are supposed to get them are supposed to first validate at the highest speed, then move the multiplier downward to validate lower speeds.
When the engineering samples are not rated very highly (the most recent example being the 2.8GHz Prescott engineering samples), that usually means there’s ramping trouble.
Putting out a 2.2GHz chip rather than a 2.6GHz chip would see to indicate that AMD is having a bit of a problem putting together a chip that will run at 2.6GHz under default conditions. They aren’t all that far away, but they’re probably not ready yet.
This would tend to confirm the real reason why FXs aren’t going to 90nm anytime soon, and faster 90nm Athlon 64s have been delayed until early next year.
However, if nothing else, AMD has gotten far enough to at least begin to start replacing current 130nm chips with (hopefully) cheaper 90nm chips.