Hello XDNA!

We're documenting how to program AMD's NPUs in Ryzen AI chips.

Our website covers the ISA, register files, operation latencies, and hand-optimized assembly kernels for tensor contractions.

Measured single-compute-tile throughput:
• XDNA1 (Ryzen 7 8700G): 398 BF16 GFLOPS (86% of peak)
• XDNA2 (Ryzen AI Max PRO 390): 1760 BFP16 GFLOPS (95% of peak)

https://tnzr.org/xdna

#AMD #RyzenAI #NPU #XDNA #XDNA2 #AssemblyLanguage #VLIW

Overview — Hello XDNA

#AMD Just Showed Off Its Own DGX Spark Rival, The Ryzen #AIHalo: Full #ROCm Support, Day-0 Support For Leading #AI Models, Available In Q2
AMD Ryzen AI MAX family, codenamed #StrixHalo, has seen adoption, from laptops to handhelds and Mini PC; every consumer PC segment. High-performance and premium SoCs offer amazing performance thanks to #Zen5 #CPU, #RDNA 3.5 #GPU, & #XDNA 2 #NPU architectures. AMD developed its own Mini PC featuring these chips, called the #RyzenAIHalo.
https://wccftech.com/amd-showed-off-its-own-dgx-spark-rival-the-ryzen-ai-halo-full-rocm-support-q2-launch/

I wonder what #AMD will call their next gen #Radeon. I hope they'll not pull an Intel and use 5 digits.

Radeon X500 series might make sense, because it's gonna be #XDNA 5, right?

Also, adding more Xes to product names increases performance by at least as much as RGB - everyone knows that.

There are #Framework13 models now with #AMD CPUs! This is getting interesting for many reasons! 👀

1. The AMD Ryzen AI CPUs come with a #NPU based on #XDNA from the #Xilinx acquisition. In practice #AMDXDNA only made it into Linux 6.14 (e.g. in the very recent Fedora 42) and there is no end user support as of today making the NPU dead hardware as of today and probably most of 2025.

2. The AMD Ryzen CPUs come with AVX512 and that's working end to end unlike whatever the f #Intel is doing there.

AMD XDNA Linux Driver v3 Published For Ryzen AI Upstreaming

AMD XDNA Linux Kernel Driver For Ryzen AI Updated

AMD Introduces Amuse 2.0 Beta with AMD XDNA Super Resolution: a fully local, AI experience

AMD Blogs: AI image generation is one of the most popular use-cases for generative AI and a quintessential AI PC experience is simply incomplete without it. Most of this local image generation happens using Stable Diffusion models created by Stability AI – which is a pioneer in this field...

Windows 11 Forum
Watch "AMD at Computex 2024: AMD AI and High-Performance Computing with Dr. Lisa Su" on YouTube
https://www.youtube.com/live/MCi8jgALPYA?si=stEWHChgymtbaCdM
> The Future of High-Performance Computing in the AI EraJoin us as Dr. Lisa Su delivers the Computex 2024 opening keynote
#AMD #Ryzen #Ryzen9000 #AM5 #Zen5 #Computex #AI #NPU #Strix #XDNA #radeon
AMD at Computex 2024: AMD AI and High-Performance Computing with Dr. Lisa Su

YouTube

My honest thought?

Xbox is waiting to cook something with AMD and XDNA2, their proprietary AI/ML accelerator unit.

XDNA2 comes out commercially this year. Microsoft should provide access through DirectML.

I think Sony just modified XDNA1 and rebranded it for graphics.

https://www.youtube.com/watch?v=SWp51cuFT3E

#Videogames #Gaming #AMD #Sony #PlayStation5 #Xbox #Microsoft #Ryzen #XDNA #AI #ML #MachineLearning #ArtificialIntelligence

PS5 Pro: How Did Sony Beat Microsoft In Machine Learning For Consoles?

YouTube
#AMD has published a blog where it provides a setup guide on how to utilize its hardware to run your very own localized #chatbot powered by #GPT-based #LLM (#LargeLanguageModel) on a diverse range of hardware such as the #Ryzen #AI PCs which include #Ryzen7000 & #Ryzen8000 #APU featuring the #XDNA #NPU along with the latest #Radeon7000 #GPU which feature AI accelerator cores.
https://community.amd.com/t5/ai/how-to-run-a-large-language-model-llm-on-your-amd-ryzen-ai-pc-or/ba-p/670709
How to run a Large Language Model (LLM) on your AMD Ryzen™ AI PC or Radeon Graphics Card

Did you know that you can run your very own instance of a GPT based LLM-powered AI chatbot on your Ryzen™ AI PC or Radeon™ 7000 series graphics card? AI assistants are quickly becoming essential resources to help increase productivity, efficiency or even brainstorm for ideas. Not only does the local...

AMD.com