Загрузка видео...
Не удалось загрузить видео
AI is not just changing what CPUs need to do. It’s changing how they are designed. AMD Senior VP, Corporate Fellow and Chief Architect of AMD CPUs Mike Clark shares how engineers are using AI to explore more design possibilities, accelerate verification and narrow down options faster. Learn how... show more
35,219 просмотров • 3 дней назад •via X (Twitter)
Комментарии: 13

Full circle: AI helping design AMD silicon, and AMD silicon running my AI. Luna, my 100% local French assistant, runs qwen3:14b on an RX 9070 XT via llama.cpp + ROCm. No cloud, no API key, 0.51s per tool call. #LocalLLM #ROCm #AMD

Great point!

amd talking about ai in chip design is funny because half their instinct gpus ship to companies training the models that will help design the next instinct gpu. tight little loop.

all changes when the photon cpu comes out mainstream. What could bottleneck that?

AI 探索更多设计可能性是吧,验证工程师的发际线先替你们谢谢 AI——芯片设计里验证本来就是最烧时间的环节,设计得越多,仿真 farm 烧得越旺。

🙄😳😳🤨 omg

verification bottlenecks in EDA tooling are getting brutal

The real unlock is turning a larger design space into better decisions, not just more options. When exploration and verification stay in the same loop, expertise gets more leverage instead of getting buried in repetition.

Facts!

verification eats 60-70% of chip design cost, so ai narrowing bug space is where the real dollars move. floorplanning and place-and-route with reinforcement learning has been shipping quietly since google's alphachip paper, amd is just now saying the quiet part loud. next-gen zen tapes will probably compress from 4 years to 2.5.

AI-driven design is powerful, but if the AI only narrows options from known spaces, how do you ensure it doesn't miss radical architectures outside that data?

ai designing the chips that run ai, we're so back

AMD’s using AI to optimize die layouts. It spots patterns humans miss, saving silicon and time. When the tools build themselves smarter, that extra efficiency is just free horsepower for whatever we’re running next. Nice work

















