Back to home
Technology

AMD Fluid Motion Frames 3 spotted in the upcoming AMD Adrenalin driver branch — could lean on AI model used in FSR 4

Source

Tom's Hardware

Published

TL;DR

AI Generated

AMD is developing Fluid Motion Frames 3 (AFMF 3) as revealed in upcoming AMD Adrenalin 25.20 driver branch notes. The driver focuses on AI with updates like Python 3.12 support and PyTorch to enhance LLM functionality on specific GPUs and APUs. AFMF 3's details are not official, but it may incorporate the same AI model as FSR 4, potentially exclusive to RX 9000 GPUs. AFMF is a driver-based alternative to FSR for games lacking FSR support, but its frame generation quality is inferior to FSR's.

Read Full Article

Similar Articles

NVIDIA CEO is the AI GPU Godfather: Amazon and Google tell Jensen when they're making AI chips

NVIDIA CEO is the AI GPU Godfather: Amazon and Google tell Jensen when they're making AI chips

NVIDIA's CEO, Jensen Huang, is considered the "AI GPU Godfather" with companies like Amazon and Google informing him when they are developing AI chips. This relationship highlights NVIDIA's prominent role in the AI chip market. The article discusses the significance of NVIDIA's leadership in AI technology and its collaborations with major tech companies.

TweakTown
Intel could be working on its own multi-frame generation tech — XeSS MFG name and logo found in Arc graphics driver files

Intel could be working on its own multi-frame generation tech — XeSS MFG name and logo found in Arc graphics driver files

Intel may be developing its own multi-frame generation (MFG) technology, potentially named XeSS MFG, as hinted in code found in its Arc graphics driver files. This move could be a response to Nvidia's MFG technology showcased in the RTX 50 series. The discovery of a logo associated with XeSS MFG suggests Intel is working on this feature for its graphics cards. While this finding is speculative, if confirmed, it could give Intel a competitive edge against Nvidia and AMD in the GPU market. Despite recent concerns about Intel's Arc lineup following a deal with Nvidia, the company remains committed to its graphics project.

Tom's Hardware
China's latest GPU arrives with claims of CUDA compatibility and RT support — Fenghua No.3 also boasts 112GB+ of HBM memory for AI

China's latest GPU arrives with claims of CUDA compatibility and RT support — Fenghua No.3 also boasts 112GB+ of HBM memory for AI

Innosilicon Technology introduces the Fenghua No.3 GPU, boasting compatibility with Nvidia's CUDA platform and support for ray tracing. The graphics card, based on RISC-V architecture, offers over 112GB of HBM memory, making it suitable for AI applications. It is marketed as an all-function GPU for various sectors like AI, scientific computing, CAD, medical imaging, and gaming. The Fenghua No.3 supports DirectX 12, Vulkan 1.2, and OpenGL 4.6, with claims of smooth gameplay in titles like Tomb Raider and Valorant. Additionally, it is China's first GPU to support YUV444 format, ideal for CAD and video editing, and can drive up to six 8K monitors at 30 Hz.

Tom's Hardware
FSR 4 modded to run on RDNA 2 GPUs improves image quality by "leaps and bounds," but carries 10% performance penalty —AMD's leaked source code turns into modding frenzy

FSR 4 modded to run on RDNA 2 GPUs improves image quality by "leaps and bounds," but carries 10% performance penalty —AMD's leaked source code turns into modding frenzy

AMD's accidental open-sourcing of FSR 4 led to modders adapting it for RDNA 2 GPUs, significantly enhancing image quality at the cost of a 10% performance hit. The modding frenzy extended to Windows, with users reporting improved visuals but reduced FPS on various GPUs. Despite performance penalties, FSR 4 offers better image quality, prompting users to adopt it for the visual upgrade. AMD is developing FSR Redstone for broader GPU compatibility, while modders continue to tweak FSR 4 for improved performance and fidelity.

Tom's Hardware

We use cookies

We use cookies to ensure you get the best experience on our website. For more information on how we use cookies, please see our cookie policy.