With the launch of the RX 480, AMD has kicked off their new RX 400 series of graphics cards. Utilising a new naming scheme, the GCN 4.0 based GPUs are set to continue the GCN legacy on the new 14nm process node. To help us better understand where all the GPUs fall in the lineup, AMD has released a decoder for their product names.
To start it off, all graphics cards with more than 1.5 TFLOPs of performance and 100GB/s of memory bandwidth get the RX branding instead of just the 4xx. For the Tier 5 cards, there will be a bit of mixed branding with the RX 460 and a plain old 460. Given the information provided for Polaris 11 which is the RX 460, we can expect the 460 to feature around 896 shaders and have low clocked GDDR5.
The lineup is further divided into different segments meant for different markets. At the top, we have the RX 490 which reportedly are meant for 4K and feature memory buses greater than 256bits. This suggests that AMD is either going with HBM2 or a 384bit bus and they’re not using GDDR5X. It also leaves open the possibility that Vega 10 will be the RX 490 since Vega features the extra-wide HBM2.
Finally, AMD has also prepared their naming scheme for mid-generation refreshes. That is where the 4×5 names come in, which are faster variants of the same GPU. We might see these when the process matures, yields get better or AMD simply pushes higher clocks. This will be like the 7950/7950 Boost and 7970/7970 GHz Ed.
I’m quite certain its already been confirmed that the RX 460 has 16 CUs and therefore 1024SPs, not 896 like Bonaire did.