News

So crazy! AMD wants to improve FSR and generate 8 times more frames: surpassing NVIDIA and Intel

2 min read
Kuai Technology reported on July 12 that recently some users have unearthed hidden multi-frame generation options in the Radeon driver, which supports up to 8x frame insertion rate. There are also overlay switches such as FSR Ray Regeneration (light regeneration) and Neural Radiance (neural radiation), suggesting that AMD is brewing a large-scale FSR technology upgrade. The current frame generation function of AMD FSR only inserts one frame between two rendered frames, which is consistent with the idea of ​​​​NVIDIA DLSS 3, while multi-frame generation allows batch insertion of frames at a higher magnification. NVIDIA supports 4x multi-frame generation when DLSS 4 is first released on the RTX 50 series, that is, three AI frames are generated for each rendered frame. Intel's multi-frame generation also supports up to 4x. The latest DLSS 4.5 version of NVIDIA is further expanded to 6x and introduces dynamic multi-frame generation, which can automatically switch the frame insertion rate according to the GPU load. If the 8x option that appears in the AMD driver is true, it will surpass the current highest specification of NVIDIA in frame insertion rate. In addition to the MFG magnification option, the driver also has overlay switches for FSR Ray Regeneration and Neural Radiance. Ray Regeneration is AMD's real-time denoiser based on machine learning. It can clean up reflection noise in ray tracing scenes before performing super-resolution and frame interpolation processing to improve the final image quality. The current version has been iterated to 1.2 and will be released with FSR SDK 2.3. From the naming, it is speculated that Neural Radiance may be related to neural rendering or radiometric calculation. These options exist in the form of overlays, which means that users may be able to force enable or adjust at the driver level without having to wait for the game to adapt one by one. It is worth noting that multi-frame generation does not come without costs. The higher the frame insertion magnification, the risk of input delay and visual artifacts also increases. NVIDIA compresses delay through Reflex technology in DLSS 4. It is not clear whether AMD has a corresponding delay compensation solution.