{"slug": "architecting-for-hw-diversity-meta-s-e2e-co-design-for-ai-training-on-amd-gpus", "title": "Architecting for HW Diversity: Meta's E2E Co-Design for AI Training on AMD GPUs", "summary": "Meta reported that hardware-model co-design and an analytical performance model delivered up to 1.5x improvement in end-to-end training throughput on AMD Instinct MI350X GPUs versus a directly ported baseline, in a paper submitted to arXiv on 15 October 2026. The paper, which Meta calls the first hyperscaler report on productionizing AMD GPUs for large-scale training, used recommendation model training on MI350X versus NVIDIA B200 as its case study and cites the MI350X's 1.6x HBM capacity relative to B200. Meta said its contributions were upstreamed to the open-source PyTorch, Triton, and ROCm ecosystems, and that agentic workflows lowered the barrier to porting from the CUDA ecosystem for microarchitecture-aware kernel optimizations and custom shim generation.", "body_md": "# Computer Science > Distributed, Parallel, and Cluster Computing\n\n  [Submitted on 15 Oct 2026]\n\n# Title:Architecting for Hardware Diversity: Meta's End-to-End Co-Design for High-Performance AI Training on AMD GPUs\n\n[View PDF](https://arxiv.org/pdf/2610.22230)\n\n[HTML (experimental)](https://arxiv.org/html/2610.22230v1)\n\nAbstract:Meta deploys a heterogeneous fleet comprising NVIDIA GPUs, AMD GPUs, and proprietary MTIA accelerators to reduce costs and mitigate vendor lock-in. However, the extensive history of NVIDIA-specific optimizations hinders performance portability to other hardware. This paper presents our methods for optimizing alternative accelerators, using AMD Instinct MI350X versus NVIDIA B200 for recommendation model training as a case study. We propose three reusable principles. First, hardware-model co-design is essential: rather than directly porting models from NVIDIA GPUs, revising model structures and software recipes allows full exploitation of accelerator-specific strengths (e.g., MI350X's 1.6x HBM capacity relative to B200). Second, our analytical performance model enables zero-warmup optimal kernel selection across diverse datatypes and matrix shapes in production models, drastically simplifying tuning. Third, increasingly mature agentic workflows have lowered the barrier to porting from the CUDA ecosystem; we extend these workflows for microarchitecture-aware kernel optimizations and custom shim generation. Compared to the directly ported baseline, these methods deliver up to 1.5x improvement in end-to-end training throughput. To our knowledge, this is the first hyperscaler report on productionizing AMD GPUs for large-scale training. Our contributions have been upstreamed to the open-source PyTorch, Triton, and ROCm ecosystems.\n    \n\n### Additional Features\n\n### Current browse context:\n\ncs.DC\n\n### References & Citations\n\nLoading...\n\n# Bibliographic and Citation Tools\n\nBibliographic Explorer \n\n*(*[What is the Explorer?](https://info.arxiv.org/labs/showcase.html#arxiv-bibliographic-explorer))\nConnected Papers \n\n*(*[What is Connected Papers?](https://www.connectedpapers.com/about))\nLitmaps \n\n*(*[What is Litmaps?](https://www.litmaps.co/))\nscite Smart Citations \n\n*(*[What are Smart Citations?](https://www.scite.ai/))\n# Code, Data and Media Associated with this Article\n\nalphaXiv \n\n*(*[What is alphaXiv?](https://alphaxiv.org/))\nCatalyzeX Code Finder for Papers \n\n*(*[What is CatalyzeX?](https://www.catalyzex.com))\nDagsHub \n\n*(*[What is DagsHub?](https://dagshub.com/))\nGotit.pub \n\n*(*[What is GotitPub?](http://gotit.pub/faq))\nHugging Face \n\n*(*[What is Huggingface?](https://huggingface.co/huggingface))\nScienceCast \n\n*(*[What is ScienceCast?](https://sciencecast.org/welcome))\n# Demos\n\n# Recommenders and Search Tools\n\nInfluence Flower \n\n*(*[What are Influence Flowers?](https://influencemap.cmlab.dev/))\nCORE Recommender \n\n*(*[What is CORE?](https://core.ac.uk/services/recommender))\n# arXivLabs: experimental projects with community collaborators\n\narXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.\n\nBoth individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.\n\nHave an idea for a project that will add value for arXiv's community? [**Learn more about arXivLabs**](https://info.arxiv.org/labs/index.html).", "url": "https://wpnews.pro/news/architecting-for-hw-diversity-meta-s-e2e-co-design-for-ai-training-on-amd-gpus", "canonical_source": "https://arxiv.org/abs/2610.22230", "published_at": "2026-10-19 01:55:50+00:00", "updated_at": "2026-10-19 02:18:46.329825+00:00", "lang": "en", "topics": ["ai-infrastructure", "ai-chips", "machine-learning", "ai-research", "developer-tools"], "entities": ["Meta", "AMD", "AMD Instinct MI350X", "NVIDIA B200", "NVIDIA", "MTIA", "PyTorch", "ROCm"], "also_reported_by": [], "alternates": {"html": "https://wpnews.pro/news/architecting-for-hw-diversity-meta-s-e2e-co-design-for-ai-training-on-amd-gpus", "markdown": "https://wpnews.pro/news/architecting-for-hw-diversity-meta-s-e2e-co-design-for-ai-training-on-amd-gpus.md", "text": "https://wpnews.pro/news/architecting-for-hw-diversity-meta-s-e2e-co-design-for-ai-training-on-amd-gpus.txt", "jsonld": "https://wpnews.pro/news/architecting-for-hw-diversity-meta-s-e2e-co-design-for-ai-training-on-amd-gpus.jsonld"}}