Navi Dicussion Thread

[SteelCrysis]

https://www.tomshardware.com/news/amd-rx...ram-pcie-4

German website pcgameshardware.de recently tested both RX 5500 XT versions using PCIe 3.0 and PCIe 4.0. The results were a bit surprising. They found that when the 4GB card's VRAM buffer was full, running on a PCIe 4.0 slot improved performance by reducing the impact of using VRAM. The 8GB also displayed minor improvements from PCIe 3.0 to 4.0, but the 8GB capacity is what makes that card shine, since the extra memory means the card didn't have to dump data off of the faster VRAM as often.

So why does this happen? Once the VRAM is filled up, the extra data is sent to the system RAM across the PCIe bus. The faster the bus, the faster the swapped data will get where it needs to go and not delay the next frame. The 5500 XT is wired as a PCIe 4.0 x8 card (physical tab is x16), which yields the same theoretical bandwidth as PCIe 3.0 x16. According to the article, AMD said that is sufficient for a GPU of this performance class. For all intents and purposes, this is correct. However, for games that eclipse the installed VRAM we see it can fall short.
...
In the end, it's a curious choice for AMD to wire these cards in an x8 configuration, particularly with the 4GB card, which has a greater chance of running into VRAM limits. While many titles will not run into these issues, those that do can take a severe performance hit.

With all of this information, I’m left wondering why AMD chose to configure the card in this manner? Cost perhaps? Increasing the performance of both cards can only help the optics, especially on the 4GB card, which is the most affected. This could have been spun as a reason to buy into the AMD Ryzen 3000 X570 ecosystem as a bright spot for PCIe 4.0. But instead, we are left wondering, why?