Since the transition to TSMC’s 20nm node does not appear to be in the cards for AMD or Nvidia, both outfits are in the process of refreshing their GPU line-ups with new 28nm parts. Nvidia is off to a good start with Maxwell and now it’s AMD’s turn.

According to Italian tech site Bits and Chips,  AMD simply needs something to fill the gap between TSMC’s planar 20nm process and upcoming FinFET processes. The 28SHP node, used for AMD’s Kaveri APU, appears to be the alternative.

The node should offer higher frequencies at lower voltages compared to TSMC’s 28nm node, even a somewhat smaller die area due to higher density. In theory, this would allow AMD to produce a more efficient GPU design, which is good news as Nvidia currently has a big lead in performance-per-watt with its Maxwell 2.0 products. A more efficient node could allow AMD to close the gap.

However, this does not mean that TSMC is off the table as a long-term partner for AMD. The company may be forced to use TSMC’s 16nm FinFET (16FF) process for future products, as the node is likely to become available ahead of Samsung’s and GloFo’s 14nm FinFET node. In addition, the 14nm FinFET node appears to be almost completely booked for mobile SoCs.