-New computing unit design: Improve efficiency and improve IPC

-Multi-level cache consistency: Reduce latency, increase bandwidth, and save power

-Streamlined graphics pipeline: Optimize unit frequency performance and increase frequency

In addition, it is the first consumer-grade GPU architecture natively supporting PCIe 4.0.

According to the information given by the hardware exposure God APISAK,The core package area of ​​the Navi GPU is only slightly larger than the three-generation Acer. The GPU itself is only 255 square millimeters, which is only 10% larger than the Polaris 10 core (232 square millimeters) used in the RX 480/580. Even almost exactly the same as the popular HD 4870 RV770 core (256 square millimeters).

In contrast,The core area of ​​the TU106 used in the RTX 2070 is 445 square millimeters, Navi is only about 57% of it, and even the TU116 core of the GTX 1660 series has 284 square millimeters, and the Navi 10 is 10% smaller.

This shows the compactness and efficiency of the Navi architecture, as well as the high integration of TSMC's 7nm process – the Turing series is still 12nm.

However, according to foreign media,Navi is not completely refurbished and rebuilt, but also follows the design of some GCN architectures.For example, each computing unit is still 64 stream processors.

It is reported that,The Navi 20, which will be launched next year for high-end enthusiasts, will be the first truly new GPU architecture for AMD, and is expected to lead AMD back to the top market.

Interestingly, ASRock has demonstrated a series of future concept graphics designs in Taipei, includingHigh-end Phantom Gaming player series, mainstream Taichi Tai Chi seriesAre all threefanDesign, it seems that the latter corresponds to the RX 5700 series, the former may be prepared for the future Navi 20, bulky, and even more than 2.5 slots.