Radeon RX 570 is a graphics card from AMD, Its 1240 MHz Boost frequency is supposed to balance the heat and performance, 1170 MHz GPU frequency is not really high, but now the core frequency is not an important parameter to determine the performance of the graphics card, The graphics processor used by the GPU is Polaris 20, Its number of 5.7 billiontransistors has a small gap compared to other brands of the same level of graphics cards. It is neither a leader nor a backward, Although the memory frequency of 1.7 Gbps is not high, it has a certain additive effect on the data processing speed, The entire graphics card layout uses the GCN 4.0 architecture, Its number of 5.7 billiontransistors has a small gap compared to other brands of the same level of graphics cards. It is neither a leader nor a backward, The stream processor is a very important parameter, responsible for coloring and rendering work. Radeon RX 570 stream processing number is 2048, and I believe its coloring and rendering speed should not be slow, The core chip process technology is 14 nm, and every process technology improvement is a huge improvement, 256 bit bit of memory interface, allowing it to instantly through put data, Radeon RX 570 has a number of texture units of 128 that can handle images very finely. That is, the image is not jagged, Although 232 mm² chip area is not large, with the current,AMD process technology, you can put enough transistors, 4 GB GDDR5 video memory doesn't look too big, but it's enough to handle most games, The maximum power is 120 W (the higher the power, the greater the heat), In order to stably perform higher performance, additional power supply is required, and the power supply interface specification is 6 Pin, It has a good single-precision floating-point computing power of 5.0 TFLOPS.

Radeon RX 570, using the GCN 4.0 architecture.

The GCN 4.0 architecture Polaris 20 chip consists of the following main components:

1. Based on Samsung's 14nm FinFET process, it delivers 15% performance improvement and up to 2.8 times power consumption per CU unit than the previous 28nm process.

2. The same macro parallel structure as “Fiji”, 36 groups of CU units are divided into 4 Shader Engines, and each Shader Engines contains 9 groups of CUs. The Shader efficiency of each CU has increased by 15% compared to the previous generation, and the geometry engine has also improved.

3. Four Geometry Units corresponding to Shader Engines and four sets of Rasterizers.

4. The total number of ACE asynchronous computing units is 4 groups. At the same time, there are 2 new sets of HWS designed for asynchronous operations are added in the architecture. HWS is used to better schedule hardware resources and improve the efficiency of asynchronous operations.

5. The L2 cache is twice as large as before, which improves the power efficiency while reducing the memory bandwidth requirement, and is more energy efficient in the memory operation.