Optimize productivity by meeting next-level demands of the AI era
summary
AMD and Microsoft describe an AI stack centered on the AMD Instinct MI300X with AMD ROCm software and OpenAI Triton support, claiming 1.5 TB total memory per server and 2.4× server memory capacity versus the market leader, plus 60% more memory bandwidth and ~30% higher peak FLOPS. Azure reports nightly regression on 700,000 hugging face models, Day‑0 PyTorch upstreaming, public VM availability, and measured leading price performance on GPT‑4 for inferencing at cloud scale.
AI-written summary · about
chapters (10)
- 0:00AI era productivity demands
- 1:16AI technology impact
- 3:47Compute needs for AI models
- 5:56AMD MI300X strategy
- 8:53Hardware and software support
- 10:02Microsoft’s AMD journey
- 12:13MI300X availability
- 14:27System optimization with Microsoft
- 17:23Customer adoption examples
- 19:33Q&A
related talks
22mAI @Scale
05mBuild AI Anywhere with ROCm™ Software: AMD and PyTorch Bring...
46mAMD + Azure: Outstanding performance and the latest security features
03mKeynote: AMD & PyTorch: A Powerful Combination for Generative AI
22mCurrent State of AI
12mAI-powered infrastructure:
58mKeynote: Enabling the AI Revolution: Next-Generation Memory Solutions for AI Computing(…)
28mHarnessing generative AI with NVIDIA AI and Microsoft Azure