TechPowerUp revealed new specifications of two upcoming AMD GPUs based on the 4th generation GCN “Polaris” architecture.
Suppose that each CU still consists of 64 stream processors (SP), the total SP count will be 2,048. This chip is supposedly going to offer a single-precision floating point performance of 5.5 TFLOP/s. The “Hawaii”/”Grenada” based FirePro W9100 which launched earlier this February had only 5.2 TFLOP/s and that SKU featured all 2,816 SP present on the chip. So this chip is faster than most “Hawaii” based SKUs. The efficiency rose by about 45%.
While “Hawaii” based SKUs feature TDP 275W, the new chip has a TDP rated no higher than 150W. AMD could pull off a “single 8-pin power connector” feat like NVIDIA.
The chip also features a 256-bit wide GDDR5/GDDR5X memory interface and 8 GB standard memory amount. The first SKUs based on this chip would feature 7 Gbps GDDR5 memory.
AMD will support the HVEC/H.265 hardware encode/decode acceleration, DisplayPort 1.3, and HDMI 2.0a outputs.
The smaller “Polaris” chip referred as “Baffin” is scheduled for 2016 and could feature 14 compute units, equaling to a stream processor count of 896. It will replaces the current R9 370/360 series SKUs.
This chip’s peak single-precision floating-point performance could rate at 2.5 TFLOP/s. Its TDP is just 50W, and it is expected to feature a 128-bit wide GDDR5 memory interface, holding 4 GB of memory.