When you purchase through links on our articles, Future and its syndication partners may earn a commission.

Nvidia currently dominates the GPU market, thanks to a combination of performance, features and brand recognition. Its advanced technologies based on AI (artificial intelligence) and machine learning have proven particularly powerful, and AMD hasn’t really caught up, especially in the consumer market. But the company hopes to change that very soon.

According to an article on GPUOpen, AMD’s research is currently focused on achieving real-time path tracking on RDNA GPUs via neural network solutions. Nvidia uses its own DLSS for image upscaling with AI, but DLSS means much more than “Deep Learning Super Sampling” – there is DLSS 2 scaling, Generating DLSS 3 framesAnd 3.5 spoke DLSS reconstruction. AMD’s latest research focuses on neural denoising to remove noisy images caused by using a limited number of ray samples in real-time path tracing – essentially ray reconstruction, as far as we know.

Path tracing normally uses thousands or even tens of thousands of ray calculations per pixel. It’s the gold standard for movies, which often require hours per rendered frame. This is because a scene is rendered using calculated ray bounces where even a slight shift in the path taken can result in a different pixel color. Do this often and accumulate all the resulting samples for each pixel, and eventually the quality of the result improves to an acceptable level.

To perform real-time path tracing, the number of samples per pixel must be significantly reduced. This results in more noise, as light rays often fail to reach certain pixels, leading to incomplete illumination requiring denoising. (Movies also use custom denoising algorithms, because even tens of thousands of samples don’t guarantee perfect output.)

AMD aims to solve this problem with a neural network that performs denoising while reconstructing scene details. Nvidia’s solution has been praised for its preservation of details that traditional rendering takes much longer to achieve. AMD hopes to achieve similar gains by reconstructing the traced path details with a few samples per pixel.

The innovation here is that AMD combines scaling and denoising within a single neural network. In AMD’s own words, their approach “generates high-quality denoised and oversampled images at a display resolution higher than the rendering resolution for real-time path tracing.” This unifies the process, allowing AMD’s method to replace multiple denoisers used in renderers and perform upscaling in a single pass.

This research could potentially lead to a new version of the AMD processor FSR (FidelityFX Super Resolution) this could match Nvidia’s performance and image quality standards. Nvidia’s DLSS technologies require dedicated AI hardware on RTX GPUs, as well as an optical flow accelerator for frame generation on RTX 40 series GPUs (and later).

AMD’s current GPUs generally lack AI acceleration features, or in the case of RDNA 3, there are AI accelerators that share execution resources with the GPU’s shaders, but in a more flexible manner. optimized for AI workloads. What’s unclear is whether AMD can run a neural network for denoising and scaling on existing GPUs, or whether this will require new processing clusters (i.e. tensor units). Achieving this on existing hardware would potentially allow a future FSR iteration to run on all GPUs, but it could also limit quality and other aspects of the algorithm.

We’ll have to wait and see what AMD ultimately offers. A refined approach to neural path tracing and scaling could bring accessible, high-fidelity graphics to a wider range of hardware, but given the demands of path tracing in games (see: Alan wakes up 2, Black Myth WukongAnd Cyberpunk 2077 RT Overdrive), we believe AMD will need much faster hardware than existing products to achieve higher levels of image fidelity.