Today AMD announced the Radeon RX 500 series graphics cards, based on a “refined” variants of the “Polaris10” family of GPUs that power the RX 400 series. AMD improved its 14 nm FinFET process, which enables higher clock speeds; lower idle and multi-monitor power draw.
The RX 500 series included four SKUs, beginning with the Radeon RX 550 at a sub-$80 price point, followed by the Radeon RX 560, which succeeds the RX 460 at $120; the RX 570, which starts at $169, and the RX 580, which is priced at $199 for the 4 GB variant, and $229 for the 8 GB variant.
The RX 580 and RX 570 are based on the 14 nm “Lexa” Polaris20 silicon. This chip is identical to the “Ellesmere” Polaris10, except for the manufacturing level improvements that help higher clock speeds. The RX 580 features 2,304 stream processors across 36 compute units, 144 TMUs, 32 ROPs, and 8 GB or 4 GB of GDDR5 memory across a 256-bit wide memory interface. The core is clocked at 1257 MHz, with 1340 MHz boost, and 8.00 GHz (GDDR5-effective) memory. The RX 570, on the other hand, features 2,048 stream processors across 32 compute units, 128 TMUs, 32 ROPs, and clock speeds of 1168 MHz core, 1244 MHz boost, and slightly faster 7.00 GHz (GDDR5-effective) memory. Although available in 4 GB and 8 GB variants.
Below the RX 580 and 570, is the RX 560, which features all 1,024 stream processors physically present on the “Baffin” silicon, across 16 compute units (896 stream processors and 14 CUs of the RX 460). The RX 560 comes with 1175 MHz core, with 1275 MHz boost clock speeds, and 7.00 GHz (GDDR5-effective) memory. It features 4 GB of GDDR5 memory across a 128-bit wide memory interface. Finally, the RX 550 is an entry-level IGP replacement. Based on the smaller Polaris12 silicon, this chip features 512 stream processors across 8 CUs, and 2 GB of GDDR5 memory across a 128-bit wide memory interface. The RX 550 clock speeds is 1183 MHz.