Both GPUs are the first products of the Radeon RX 5000 Family, which use the GPU chip "Navi" with 7nm Fabrication technology. The model number '5700' on this product may imply that they will not occupy the top position in the 'Navi' lineup, though it may cause some confusion over the popular RX 570 'Polaris' because it offers amazing price vs. performance.
Navi 'itself is not AMD's first 7nm GPU, because some time before, AMD released Radeon VII based on' Vega '.
As with all the latest GPUs, specifications are the most awaited enthusiasts. Let's look at the specifications of the two Radeon RX 5700 XT.
Of the two Navi in circulation, the Radeon RX 5700 XT is the highest version. This GPU is equipped with 40 CU (2560 Stream Processor), 160 Texture Units, and 8GB GDDR6 which has a rating of 14Gbps. With 256-bit memory bus width, it has an effective bandwidth of 448 GB / s. Although it does not come with HBM2 memory like Radeon VII, the bandwidth of GDDR6 seems to be quite competitive, even nearly equivalent to HBM1 Bandwidth, 512 GB / s on the Radeon R9 Fury X.
AMD provides 64 ROPs on the RX 5700, exactly 2x more than the mainstream Polaris GPU configuration in 32.
The speed / clock speed of the Radeon RX 5700 XT seems to adjust to the load, where AMD provides 1605 Mhz Base Clock, 1755 Mhz Game Clock and up to 1905 Mhz Peak Boost Clock (Briefly, 'Game' Clock is the GPU speed range when experiencing normal game load , not heavy benchmark load, or heavy compute load).
At its maximum Boost clock, AMD claims that the 5700 XT has a 9.75 TFLOP / s single-precision compute capability.
Die Radeon 5700 XT has a relatively small size, 251 mm2, and has 10.3 billion transistors. This high density seems to be an advantage of the 7nm TSMC fabrication that is seen immediately. Later, the GPU uses PCIe 4.0 16x interconnection (Don't worry, this standard will be backward-compatible to PCIe 3.0 on most modern motherboards).
Other than those mentioned above, aside from the 40 Compute Units / 2560 Stream Processors in the Radeon RX 5700 XT, there is no other dedicated processing block to accelerate the calculation of AI / real-time ray tracing as owned by NVIDIA GeForce RTX 'Turing'.
