New details for polaris 10 and polaris 11

New details for Polaris 10 and Polaris 11 show great effort to improve efficiency

New details for Polaris 10 and Polaris 11. AMD Polaris maintains the same 64 stream processor structure for each Compute Unit (CU) as previous generations of GCN. Speaking of the total number of CUs in Polaris we find that the Polaris 11 silicon with the “ Baffin ” moniker will have a total of 1, 024 stream processors spread over 16 CUs while Polaris 10 “ Ellesmere ” will have 2, 304 stream processors spread over 36 CUs. Later the Vega architecture would arrive with a maximum of 4, 096 stream processor in 64 CU, the same configuration as the current AMD Fiji GPU.

Polaris 11 and Polaris 10 will come with GDDR5 memory to succeed Curacao and Tonga respectively, so their performance should be below Fiji-based Fury cards, although their energy efficiency will be much higher and will allow high-performance systems with very tight consumption.. Vega would arrive in 2017 to succeed Fiji in the highest range and would mark the premiere of the HBM2 memory.

With this it seems quite clear that AMD is focusing a lot on greatly increasing the energy efficiency of its new GPUs so we should not expect a large increase in gross power.

We leave you a selection of our articles on AMD Polaris:

AMD Polaris Announced, New GCN 4.0 Graphics Architecture

AMD Polaris virtual reality with passive cooling

AMD Polaris would have no HBM2 memory

AMD will have its own interconnector with Polaris

Source: techpowerup