When analyzing AMD and the OEM's for their treatment of Carrizo, I think many people are forgetting that there might be a more complex picture.
It's easy to understand what AMD needs, profit. However, equally important is something far less sexy, cost control. The last time AMD gambled big in 2012,
they lost billions. Today the market has shrunk since 2012, and the competition (Intel) has a greater node-advancement (14nm vs 28nm) than it was back then (22nm vs 32 nm). So I think AMD is holding out farely well.
We are also standing in the doorstep to the 14nm-era. What AMD does not need in that era, is a
bunch of obsolete inventory. Having some 28nm inventory even when 14nm products have launched in the market is inevitable, but it is in AMD's hands to control what that inventory will consist of.
Therefore I guess it is in AMD's best interest to get rid of their oldest products ASAP: Hawaii (inefficient), Puma+ (old, DDR3-limited) and some of their smallest GPU's (
Oland, Pitcairn).
Carrizo and Fiji is still farily new and will be able to be upgraded in the next years (DDR4 and larger HBM), so even with 14nm products on the market AMD might still be able to sell these parts. But who in their right mind would buy a Puma+ (Carrizo-L) PC or a GCN 1.0/1.1 GPU when the next gen parts have arrived? D:
Therefore, I think what we are seeing right now (Carrizo being "sabotaged" with single channel DDR3 and 15w, and Fiji sales being cannibalised by the extremely price agressive R9 390) is AMD's and the OEM's attempt to clear the channel of soon-to-be-obsolete products that won't sell in one year.