AMD senior executives, including Radeon Technologies Group head David Wang, recently spoke with 4Gamer about the Red team’s future GPU strategy, which includes the RDNA4 graphics architecture. AMD’s RDNA 3 graphics architecture was introduced and released to the public in the fourth quarter of 2022.
The RDNA 3 architecture has since been used in desktop, mobile, and integrated graphics solutions such as the Radeon RX 7900 (Desktop), Radeon RX 7000S/M (Laptop), and Radeon 700M series (iGPUs).
Beginning with Mesh Shaders, David Wang claims that AMD’s current RDDNA 3 GPUs include MDIA, a Multi-Draw Indirect Accelerator capable of processing MDI on a hardware level. This enables the company to achieve up to 2.3x the performance of RDNA 2 GPUs. It’s been proposed as a genuine replacement for the original LoD.
AMD wants to see even more advanced shaders and techniques installed within RDNA4 GPUs
David Wang also discussed the use of AI in Radeon-class GPUs for gamers. It is made clear that AI processing is mostly done on CPUs, and they are capable of doing so quickly. In fact, 95% of inference processing is said to be done on CPUs, and AMD is well aware of this as one of the industry leaders in server CPUs. Meanwhile, the AI Learning and Training model is primarily done on the GPU side, and Wang claims that their recent commitment to the ROCm software suite is now on par with NVIDIA’s CUDA platform.
In terms of the performance of the inferencing accelerator on the Radeon RX 7900 “RDNA 3” GPUs, David Wang claims that putting an AI accelerator on the GPU is more of a business decision based on what users want and don’t want. AMD does not want the AI accelerator to be limited to just upscaling features like FSR, XeSS, and DLSS. It should be noted that AMD is already working on the third generation of FSR, known as FSR 3.
According to AMD, some other applications of AI accelerators within GPUs could be NPC behaviour and movement, with the learning model being used to make those better and more fluid than the developers’ standard AI programming model. Developers can use the AI accelerator to improve the gaming experience for their customers.
The AI can also be used for image processing, which is becoming more popular among AI generational tools. Wang stated that Stable Diffusion for GPUs was added to the Vulkan ML API around the same time that the Radeon RX 7900 “RDNA 3” series was announced.
Also Read:
