When there is very little left for the reviews of the new AMD RX 6000, attentive to what we are going to do here, the first data about the performance of the new generation of cards running games with Ray Tracing support begins to leak, in concrete performance of the RX 6800 XT. Let’s not forget that during the presentation on October 28, it was one of the information that AMD did not reveal.
One of the novelties in the RX 6000 cards is the support for ray traced games , thanks to the addition of the Ray Accelerator Unit , which is homologous to the RT Core that NVIDIA added in its NVIDIA RTX 2000 a couple of years ago and which has returned to include in the form of an improved version in the RTX 3000.
We knew what the performance increase was when using the Ray Accelerator Unit and when not through the internal DirectX Ray Tracing benchmark, but that is not enough information, since we need to know how the RX 6000 performs under Ray Tracing in the games.
RX 6800 XT performance under Ray Tracing according to AMD
The RX 6800 XT has been tested in five different titles:
- Battlefield V with an average rate of 70 frames per second,
- Call of Duty MW can reach 95 frames per second,
- Crysis Remastered has the second highest rate with 90 FPS ,
- Metro Exodus reaches the lowest figure of all with 67 FPS ,
- Shadow of the Tomb Raider stays at 82 FPS .,
All of them running at 1440P resolution with ultra options, paired with an AMD Ryzen 9 5900X CPU and 16 GB of DDR4-3200 memory .
We have to take into account the enormous help that the AMD Ryzen R9 5900X supposes when it comes to maintaining a high frame rate with this AMD CPU, in any case the information comes from an AMD slide so, although the information has Visions of being true, we will still have to wait to see its performance compared to the RTX 3000.
4K Ray Tracing on the AMD RX 6800 XT via AI
AMD itself revealed a few days ago that the RX 6000 range is designed for Ray Tracing at 1440p , but AMD could show 4K performance at 60 FPS with the deployment of the DirectML super-resolution algorithm , whose operation is very similar to that of of NVIDIA’s DLSS , but not knowing what the performance of this algorithm will be on a GPU that lacks Tensor Core-style units, which are the most efficient to execute this type of algorithm.
The current idea between both manufacturers is that a GPU for 4K Ray Tracing would be too complex and expensive to manufacture, at least under the 7 nm node, hence the decision to opt for solutions based on super-resolution or scaling of resolution through artificial intelligence algorithms.
We cannot forget that a scene at 4K or 2160p moves about 2.25 more pixels on the screen than a scene at 1440p, so the necessary power and the number of Compute Units in the GPU becomes enormous, which is why it is much better. right now scaling through AI, for both NVIDIA and AMD.