AMD officially unveiled its latest generation of Radeon graphics cards yesterday. Amid all of the announcements, one question being asked is how well do AMD’s new GPUs perform at ray tracing. This is a pretty big elephant in the room, as NVIDIA introduced the technology with its RTX cards. By giving them dedicated RT cores combined with Tensor cores for DLSS, it enabled the primary GPU to offload a lot of the work needed for ray racing. Without these technologies, FPS can drop greatly in games.
Ray Accelerators
We did learn that AMD has implemented its own Ray Accelerator technology. By doing computational math for the intersection of rays, it can help improve performance. It has included one RA inside each computational unit. In a way, this is akin to the RT cores used by NVIDIA. However, this is only half of the solution that we’ve already seen NVIDIA implement.
Super Sampling
With DLSS or Deep Learning Super Sampling, NVIDIA has shown how AI-based image reconstruction and upscaling can improve performance when ray tracing is enabled. AMD is working on a similar feature that is open and cross platform. By not making it a proprietary feature, it hopes to gain even greater industry support. It could also be implemented with Microsoft and Sony’s new consoles.
Proper Implementation
The biggest complaint with DLSS comes from the source resolution. Too low, and it is often blurry or muddy looking. Too high, and FPS takes a hit. Finding the sweet spot is key to super sampling. Here’s hoping that AMD’s solution enables the best performance and visual quality possible.