AM3 can't compete on 28nm, since it's "custom design"
Could you elaborate a little bit? How is this different from your own custom implementation of SHA256?
Unlike BitFury and AM, we didn't use custom circuit design (with custom PDK) in our 1st and 2nd gen.
It's a time consuming effort, that needed to be redone from the beginning when moving to a more advance process node.
We're doing that with our 3rd gen and we understand the effort. I don't believe AM can produce a competitive 28nm in time.
BitFury on the other hand will probably do.