AMD not drawn into AI feature war with NVIDIA: RDNA4 architecture will emphasize ‘game performance’ – AMD

This is what they argued In an interview they did with the Japanese website 4 jammersDavid Wang (Senior Vice President, Radeon Technologies Group.) and Rick Bergman (Executive Vice President, Computing and Graphics Business) who provided the first information on the direction AMD’s next-generation graphics architecture will take. While they acknowledge NVIDIA’s move to GPU-accelerated AI, on the other hand, they say they don’t think spending GPU-accelerated AI computing resources on image processing and performance scaling is the best.

AMD has its own technology for image processing and upscaling, FSR (FidelityFX Super Resolution) but it does not use AI acceleration. David Wang stated that with the introduction of AI accelerators with the RDNA3 architecture, he hopes to take advantage of this ability to improve various areas of gameplay — such as progressive world generation, NPCs, AI bots, etc. — so as to add a new level. It is rather complicated than spending hardware resources on image processing.

AMD has also emphasized the need to make the GPU more independent of the CPU when it comes to graphics rendering. The company has made several steps in this direction with recent generations of graphics cards, with its latest move being the introduction of a Multi-Indirect Induction Accelerator (MDIA) with the RDNA3 architecture. With this component, the program can forward multiple drawing commands to the GPU which greatly reduces the load at the CPU level. The RDNA3 architecture is 2.3 times more efficient in this area than the previous generation (RDNA2). The company will introduce further innovations in this direction using the RDNA4 architecture.

Understandably, AMD representatives haven’t addressed the “what,” “how,” or “when” regarding RDNA4 architecture graphics cards since the third generation architecture has only been on the market for a few months.