Optimized GPU Software Stack

AMD ROCm™ is an open software stack including drivers, development tools, and APIs that enable GPU programming from low-level kernel to end-user applications. ROCm is optimized for Generative AI and HPC applications, and it is easy to migrate existing code into ROCm software.

  • AMD Instinct™ Accelerators
    ROCm supports all AMD Instinct™ accelerator models.
  • AMD Radeon™ Graphics
    Certain ROCm features support select AMD Radeon™ graphics cards.
AMD Instinct™ Accelerators
ROCm supports all AMD Instinct™ accelerator models.
AMD Radeon™ Graphics
Certain ROCm features support select AMD Radeon™ graphics cards.

What’s New with ROCm 7

Full support for AMD Instinct™ MI350 Series GPUs

Distributed inference with open-source framework support

Enterprise-ready AI tools with orchestration and endpoint integration

Support for large-scale models with new data types FP6 and FP4 for enhanced flexibility and performance.

Enhanced code portability with HIP 7.0

See What’s New

ROCm Evolution Over the Years

Leading enterprises and research institutes have been leveraging ROCm for nearly a decade. Explore the various milestones that are a part of the history of ROCm.

  • ROCm 1.0
    The first version of ROCm is developed. AMD demonstrates CUDA to HIP port of Caffe and Torch7 using the HIPIFY tool.
  • ROCm 2.0
    ROCm includes Linux Kernel upstream support and MIOpen deep learning libraries.
  • ROCm 3.0
    ROCm includes AMD Infinity Fabric™ technology support. RCCL communication libraries are released. The ecosystem expands to support cluster management and deployment tools (rocSMI, Docker, Kubernetes, SLURM, and more). ROCm includes upstream support for PyTorch.
  • ROCm 4.0
    ROCm supports AMD CDNA™ architecture.
  • ROCm 5.0
    PyTorch official package available. ROCm supports AMD CDNA 2 architecture.
  • Frontier System
    Oak Ridge National Laboratory’s Frontier system becomes the first official Exascale system. AMD becomes a founding member of the PyTorch foundation.
  • ROCm 6.0
    AMD partners with Hugging Face, enabling thousands of models. ROCm includes day-zero support on PyTorch 2.0. ROCm enables PyTorch AI at scale, with a 1 trillion parameter model successfully getting trained on the Frontier system. ROCm supports AMD CDNA 3 architecture.
  • GPT-4
    ROCm enables cutting-edge AI, with Microsoft’s GPT-4 now running at scale. The ecosystem grows, with OpenAI Triton support for ROCm.
  • ROCm 7.0
    Providing full support for MI350 Series GPUs and adding scalability through cluster-wide orchestration and additional enterprise capabilities.
ROCm 1.0
The first version of ROCm is developed. AMD demonstrates CUDA to HIP port of Caffe and Torch7 using the HIPIFY tool.
ROCm 2.0
ROCm includes Linux Kernel upstream support and MIOpen deep learning libraries.
ROCm 3.0
ROCm includes AMD Infinity Fabric™ technology support. RCCL communication libraries are released. The ecosystem expands to support cluster management and deployment tools (rocSMI, Docker, Kubernetes, SLURM, and more). ROCm includes upstream support for PyTorch.
ROCm 4.0
ROCm supports AMD CDNA™ architecture.
ROCm 5.0
PyTorch official package available. ROCm supports AMD CDNA 2 architecture.
Frontier System
Oak Ridge National Laboratory’s Frontier system becomes the first official Exascale system. AMD becomes a founding member of the PyTorch foundation.
ROCm 6.0
AMD partners with Hugging Face, enabling thousands of models. ROCm includes day-zero support on PyTorch 2.0. ROCm enables PyTorch AI at scale, with a 1 trillion parameter model successfully getting trained on the Frontier system. ROCm supports AMD CDNA 3 architecture.
GPT-4
ROCm enables cutting-edge AI, with Microsoft’s GPT-4 now running at scale. The ecosystem grows, with OpenAI Triton support for ROCm.
ROCm 7.0
Providing full support for MI350 Series GPUs and adding scalability through cluster-wide orchestration and additional enterprise capabilities.

Partnerships and Proven Success

Mosaic ML

See how Mosaic ML leverages ROCm to enable LLM training on AMD Instinct accelerators with zero code changes and at high performance.

Read Blog

Hugging Face

Learn about how Hugging Face and AMD partner to accelerate state-of-the-art models, and how Hugging Face is integrating ROCm in their open-source libraries.

Read Blog

PyTorch

AMD is aligning with the PyTorch Foundation on the value of an open software ecosystem for AI and is driving ROCm support for PyTorch.

Read Blog

OpenXLA

Explore more on OpenXLA—an open-source ML compiler ecosystem that Google has co-developed with AMD.

Read Blog

Addressing the Most Demanding Workloads 

ROCm enables AI and HPC application development across a broad range of demanding workloads.

AI

Large Language Models

Image and Video Generation

Recommendation

Computer Vision

HPC

Life Sciences

Computational Fluid Dynamics

Energy

Environmental Science

ROCm for AI Workloads

ROCm offers a suite of optimizations for AI workloads and supports the broader AI software ecosystem including open frameworks, models, and tools.

Discover ROCm for AI
ROCm AI Developer Hub
Simplified Model Development

Supports a wide range of models on Hugging Face  that can be leveraged to develop user-specific solutions

Support for Leading Frameworks

Provides upstream support for leading AI frameworks including TensorFlow and PyTorch

Machine Learning Libraries

Offers dedicated libraries for machine learning, including MIOpen and MIVisionX

Scale-Out Deployments

Supports leading containerization tools including Docker, Singularity, Kubernetes®, and Slurm to enable deployment at scale

Build your AI Models on the AMD Developer Cloud

Apply for access to AMD Instinct™ accelerators and streamline your development.

Apply Today

ROCm for HPC and Supercomputing

ROCm software enables HPC across a variety of disciplines to help solve some of the world’s toughest challenges. ROCm is also enabling application development on some of the world’s top supercomputers.

Discover ROCm for HPC
ROCm HPC Developer Hub
HPC Programming Languages

Supports multiple programming language choices for HPC workloads, giving developers a range of tools to help solve the problem at hand—including OpenMP®, HIP, OpenCL™, and Python™

Math and Communication Libraries

Includes a wide variety of math and communication libraries to help developers deliver more functional HPC code and drive up performance

Comprehensive Developer Tool Set

ROCm includes a comprehensive set of tools to provide developers and researchers with flexible management, quality control, and monitoring capabilities. 

Development Tools, Profilers, and Debugging

Supports a range of tools to help developers track, analyze and debug—helping reduce the time it takes to complete application development

Management Solutions for Container Deployments

Supports leading containerization tools to help simplify system and workload deployments and aid management of large-scale AMD GPU-accelerated clusters for AI and HPC

System Resource Monitoring Tools

Includes a tool for System Administrators and Cluster Managers to detect issues and troubleshoot

Remote Manager Capabilities

Includes advanced out-of-band manageability circuitry for simplified GPU monitoring in large-scale systems

Supported GPUs

AMD Instinct GPUs

Discover how AMD Instinct accelerators supercharge AI and HPC.

Discover Instinct

AMD Radeon Graphics

Certain ROCm features support select AMD Radeon graphics cards.¹ Discover next-gen professional graphics.

Discover Radeon PRO Discover Radeon RX

Developer Resources

ROCm Developer Hub

Start developing AMD GPU-accelerated applications. Visit the ROCm Developer Hub to get access to the latest user guides, containers, training videos, webinars, and more.

Get Started

AI Developer Program

Access to free AMD Developer Cloud credits, exclusive training, monthly hardware sweepstakes, and community support designed to support your AI development work.

Join Program

Infinity Hub

Access advanced software containers and deployment guides for HPC and AI applications.

Visit Infinity Hub

Documentation

Find the latest documentation on all ROCm releases.

Access Docs

Developer Blogs

Read the latest news about the ROCm features.

Read Blogs

AI Academy

Self-learning courses designed specifically for AI developers.

Learn More

ROCm Newsletter

Receive the latest ROCm news.

Stay Informed
Footnotes

©2024 Advanced Micro Devices, Inc. All rights reserved. AMD, the AMD Arrow logo, AMD ROCm, AMD Instinct, EPYC, Radeon Instinct, and combinations thereof are trademarks of Advanced Micro Devices, Inc. PyTorch is a trademark or registered trademark of PyTorch. Other product names used in this publication are for identification purposes only and may be trademarks of their respective companies.

  1. For a full list of Radeon parts supported by ROCm, go to https://rocm.docs.amd.com/en/latest/reference/gpu-arch-specs.html  
Subscribe to the latest news from AMD
© 2026 Advanced Micro Devices, Inc.