AMD Instinct™ AI & HPC Software Ecosystem

Zyphra, an open-source superintelligence company, and AMD collaborate to train frontier AI models on AMD Instinct™ GPUs through deep integration with the ROCm™ software stack. The collaboration leverages key hardware advantages of the Instinct series - large high-bandwidth memory, larger VRAM size, and high compute throughput for key datatypes - to deliver significant improvements in training speed and model efficiency.

Customers and the open-source community gain access to next-generation AI solutions and AMD benefits from the validation of its hardware and software for full-stack frontier AI model training.

Hugging Face and AMD collaborate to accelerate open and scalable AI development using AMD Instinct™ GPUs. By leveraging the AMD ROCm™ software stack, Hugging Face has enabled day-0 integration across its core libraries—including Transformers, Optimum, and Text Generation Inference—helping ensure seamless compatibility and performance optimization on AMD Instinct GPUs. Developers can train, fine-tune, and deploy millions of models from Hugging Face’s model hub with efficient scalability and high throughput.

The collaboration empowers AI practitioners to prototype and optimize NLP and generative AI applications rapidly, while demonstrating the strong synergy between the AMD open-source software ecosystem and Hugging Face’s mission to democratize state-of-the-art AI.

Essential AI and AMD collaborate to advance reasoning capabilities in large language models through cutting-edge research and infrastructure. Using AMD Instinct™ GPUs and the AMD ROCm™ software platform, Essential AI has demonstrated breakthrough discoveries in AI reasoning that emerge during pretraining phases rather than requiring extensive post-training methods.

Together, this collaboration advances our understanding of how capabilities emerge in pretraining and the development of more efficient foundation models that can achieve sophisticated reasoning with reduced computational overhead, helping democratize access to advanced AI capabilities across enterprises and research institutions.

The combination of AMD Instinct GPUs' high-performance architecture and ROCm's open development framework provides researchers with the computational foundation needed to explore new frontiers in AI reasoning.

Z.ai and AMD are collaborating to advance large language model capabilities with a focus on reasoning and multimodal intelligence. Utilizing AMD Instinct™ GPUs powered by AMD ROCm™ software, Z.ai's GLM model series benefits from AMD's open ecosystem approach to AI development, enabling flexible deployment across diverse hardware configurations.

This collaboration supports the growing demand for accessible, high-performance language models in both research and enterprise applications, while the AMD ROCm framework provides the optimization tools necessary to maximize inference efficiency and training throughput for next-generation conversational AI systems.

Stepfun and AMD are collaborating to advance multimodal AI capabilities across video generation, 3D modeling, and voice interaction technologies. Through AMD Instinct™ GPUs and AMD ROCm™ software optimization, Stepfun's Step-Video-T2V and Step-Audio models deliver state-of-the-art performance in content creation and human-computer interaction.

This collaboration enables creative professionals and enterprises to access cutting-edge video generation and voice synthesis capabilities while leveraging the AMD open software approach for flexible deployment and customization.

The integration of AMD Instinct GPU architecture with AMD ROCm's distributed inference capabilities provides the computational foundation for Stepfun's ambitious multimodal AI applications that bridge digital and physical world interactions.

Cohere and AMD have established a strategic partnership to deliver enterprise-grade language model solutions via optimized infrastructure.

Through this collaboration, Cohere deploys its Command models—including Command R+—using AMD Instinct™ GPUs and the AMD ROCm™ open software ecosystem. Command R+ is an advanced retrieval-augmented language model designed to provide high performance for enterprise use cases such as task automation, conversational AI, and complex reasoning.

By leveraging AMD Instinct GPUs and AMD ROCm supported robust AI frameworks and libraries, this partnership offers organizations flexible, secure, and scalable AI deployments across diverse enterprise environments, helping ensure efficient resource use and support for data privacy requirements.

Meta and AMD work together to optimize the performance of open-weight Llama models on AMD Instinct™ GPUs, delivering exceptional efficiency for large-scale AI workloads.

With Day 0 support and extensive tuning, developers can seamlessly deploy Llama AI models across a wide range of applications. This collaboration empowers the AI community with scalable, open-source solutions accelerated by AMD Instinct GPUs.

Liquid AI partners with AMD to support advanced AI model architectures designed for efficient training and inference on AMD Instinct™ GPUs. By combining innovative foundation model design with AMD accelerated infrastructure, organizations can deploy generative AI and reasoning systems across large scale enterprise applications.

This partnership helps deliver high-performance AI workloads while improving efficiency and scalability for the industry to harness for their own workloads.

PyTorch and AMD collaborate to deliver high-performance, accessible AI development on AMD Instinct™ GPUs through the AMD ROCm™ software stack. With upstream support in PyTorch and close co-engineering efforts, AMD helps ensure optimized performance, seamless integration, and day-zero readiness for all new PyTorch features on AMD Instinct GPUs.

This unified ecosystem enables developers to train and deploy models efficiently across AMD CDNA™ based AMD Instinct GPUs. By offering simplified installation, public Docker images, and broad model compatibility, AMD and PyTorch are jointly democratizing AI compute and accelerating innovation across research and customer AI workloads.

TensorFlow and AMD collaborate to bring high-performance AI training and inference to AMD Instinct™ GPUs via the open-source AMD ROCm™ stack.

With official ROCm-enabled TensorFlow builds and upstream integration, developers can seamlessly train and deploy models on AMD Instinct GPUs—scaling workloads and accelerating time to production. This empowers AI practitioners with greater flexibility, efficiency, and performance on AMD GPU infrastructure.

JAX and AMD collaborate to deliver high-performance AI training and inference on AMD Instinct™ GPUs through the open ROCm™ ecosystem. With upstream AMD ROCm support, official Docker images, and integration through OpenXLA and MLIR, developers can seamlessly accelerate large-scale array computations, automatic differentiation, and model execution in JAX.

This collaboration empowers researchers to explore next-generation AI workloads with exceptional throughput, memory bandwidth, and open-source flexibility on Instinct GPU–powered systems.

ONNX Runtime and AMD collaborate via the MIGraphX Execution Provider to deliver optimized, vendor-neutral AI inference on AMD Instinct™ GPUs.

Integrating MIGraphX’s graph optimizations with AMD ROCm libraries like MIOpen and rocBLAS, this collaboration accelerates ONNX models across vision, language, and multimodal workloads. Together, they enable open-standard, high-throughput inference that scales efficiently across diverse computing environments.

DeepSpeed and AMD collaborate to enable large-scale distributed training on AMD Instinct™ GPUs through AMD ROCm™ software optimization.

This collaboration supports enterprises and research institutions in training transformer models and large language models with reduced memory overhead and improved training throughput.

The integration of AMD Instinct GPUs' high-bandwidth memory with AMD ROCm distributed training capabilities enables researchers to scale model development while maintaining cost-effectiveness in large-scale AI training workflows.

MLIR and AMD collaborate through the MLIR-AIE and rocMLIR projects to deliver optimized compilation across various AMD hardware architectures. Built on the LLVM framework and integrated with the AMD ROCm™ software stack, rocMLIR provides a multi-level intermediate representation that enables advanced optimizations for convolution, GEMM, and attention operations on AMD Instinct™ GPUs and AI Engine processors.

This collaboration streamlines code generation, fusion, and model partitioning through production-tested optimization passes, empowering developers to achieve high performance and efficiency across AMD’s heterogeneous compute platforms.

OpenXLA and AMD collaborate as founding members of the open-source machine learning compiler ecosystem.

Through AMD Instinct™ GPUs and AMD ROCm™ software integration, this collaboration enables cross-platform ML infrastructure that reduces vendor lock-in and democratizes access to high-performance computing. It provides developers with unified compiler technology that bridges TensorFlow, PyTorch, and JAX frameworks across diverse hardware platforms, fostering innovation through open standards that eliminate software fragmentation in the ML ecosystem.

IREE (Intermediate Representation Execution Environment) collaborates with AMD as a founding contributor to the MLIR-based end-to-end compiler framework that targets AMD Instinct™ GPUs through ROCm™/HIP backend support.

This collaboration enables developers to deploy machine learning models across diverse AMD hardware architectures from data center accelerators to edge devices using unified intermediate representation. The collaboration provides the ML community with open-source tools for compiling models from frameworks like PyTorch, TensorFlow, and JAX to run efficiently on AMD hardware while maintaining performance portability.

Through ROCm integration and AMD contributions to the Linux Foundation AI & Data Foundation project, IREE delivers optimized code generation for AMD Instinct GPUs that supports complex ML workloads including dynamic shapes, flow control, and advanced model features.

Triton and AMD collaborate to deliver high-performance, scalable AI inference on AMD Instinct™ GPUs through deep integration with the AMD ROCm™ software stack. With full support for Triton Inference Server—including backends for vLLM and other leading frameworks—this collaboration enables seamless deployment of large language models and deep learning workloads on AMD hardware.

By combining open, cost-effective compute with optimized inference pipelines and flexible model serving, Triton and AMD empower developers and customers to accelerate production-grade AI across cloud and on-premise environments.

Makora's AI-powered platform automatically generates and tunes high-performance HIP and Triton kernels for AMD Instinct™ GPUs, eliminating the need for manual low-level GPU programming expertise.

By targeting AMD hardware directly, including AMD Instinct™ MI355X via AMD ROCm™ and rocprof-compute profiling, Makora enables developers to deploy AI models at peak performance, without specialized HIP engineering.

The platform bridges the kernel optimization gap across hardware vendors, helping customers maximize throughput and reduce inference costs on AMD Instinct GPUs while maintaining full cross-architecture portability.

Pebble and AMD collaborate to optimize LLM inference efficiency on AMD Instinct™ GPUs. Using AMD ROCm™ software and vLLM, Pebble’s Sonar platform identifies power and performance settings that improve throughput and tokens per watt for inference workloads.

Pebble reported throughput and tokens-per-watt gains in its evaluations, with work continuing toward further integration with the AMD Instinct software ecosystem.

vLLM and AMD collaborate to enable high-throughput, memory-efficient LLM inference on AMD Instinct™ GPUs.

Through AMD ROCm™ integration, advanced kernel optimizations, and support for the most advanced models, this collaboration empowers scalable, low-latency deployment of generative AI in research and production environments.

Fireworks AI and AMD collaborate to deliver high-performance AI inference on AMD Instinct™ GPUs through deep integration with the AMD ROCm™ software platform. Fireworks AI’s inference platform leverages AMD Instinct GPUs to provide low-latency, cost-efficient, and highly scalable AI model serving across cloud and enterprise environments.

Optimized with custom GPU kernels such as FireAttention for the AMD Instinct GPUs, the collaboration achieves considerable cost reduction compared to alternative implementations.

By combining parallel processing power from AMD with Fireworks’ optimized inference engine, this collaboration delivers fast, reliable, and energy-efficient AI infrastructure—validating AMD Instinct GPUs as a leading platform for next-generation generative AI deployments.

Kog AI and AMD collaborate to deliver breakthrough inference acceleration, achieving faster token generation on AMD Instinct™ GPUs compared to traditional inference engines.

Through AMD ROCm™ optimization, Kog's inference engine addresses critical latency bottlenecks in agentic workflows, voice applications, and gaming that require sequential token processing.

This collaboration enables enterprises to deploy high-speed AI applications with improved cost efficiency while leveraging AMD Instinct GPU architecture and AMD ROCm frameworks for real-time inference optimization.

Mooncake, developed by Moonshot AI, collaborates with leading LLM inference systems including vLLM and SGLang that support AMD hardware integration. Through its KVCache-centric disaggregated architecture and Transfer Engine, Mooncake enables efficient large language model serving with AMD ROCm™ compatibility.

This collaboration provides the foundation for scalable inference deployments that can leverage AMD Instinct™ GPUs for long-context scenarios, achieving significant throughput improvements while maintaining service level objectives for production AI applications.

ElioVP accelerates AI innovation within AMD ecosystem by developing optimization suites for AMD Instinct™ GPUs and AMD ROCm™ software. Through specialized Docker images and framework optimizations for popular models like DeepSeek-R1, ElioVP streamlines deployment of large-scale AI models on AMD hardware.

This collaboration provides developers with ready-to-use solutions that unlock the capabilities of cutting-edge language models on AMD accelerators, supporting both experimentation and production deployment scenarios within the AMD AI ecosystem.

Bud Ecosystem and AMD work together to bring Bud's Enterprise AI Management Platform — a single integrated platform from silicon to agents — to enterprise AI workloads running on AMD Instinct™ GPUs.

By leveraging AMD accelerated infrastructure and the AMD ROCm™ open software stack, Bud's platform delivers multi-modal inference, deployment across on-prem, hybrid, sovereign cloud, and edge environments, and unified governance and observability from a single control plane.

This collaboration enables enterprises across financial services, government, manufacturing, and retail to move from pilot to production in days, with sovereign data control, predictable unit economics, and a self-improving flywheel that compounds accuracy and efficiency with every workload.

ClearML and AMD collaborate to accelerate end-to-end enterprise AI workflows at every phase of AI/ML production on AMD Instinct™ GPUs.

ClearML's AI Infrastructure Platform, fully integrated with AMD Instinct GPUs and AMD ROCm™ software stack, is a three-layer unified solution that manages GPU clusters across environments, provides a complete AI workbench with orchestration capabilities, and enables rapid, secure deployment of AI models, apps, and agents.

Together, ClearML and AMD empower organizations to streamline the entire AI lifecycle, maximize AMD Instinct GPU utilization, and bring cost-efficient, innovative solutions to production at scale.

Rapt AI and AMD collaborate to maximize GPU utilization and accelerate AI deployment on AMD Instinct™ GPUs. Through fine-grained virtualization and real-time orchestration, Rapt AI dynamically allocates AMD Instinct GPU resources across concurrent training, fine-tuning, and inference workloads—helping ensure optimal performance, cost efficiency, and SLA compliance across cloud and on-premise environments.

Leveraging the high compute and memory bandwidth of AMD Instinct GPUs, Rapt’s intelligent workload packing and automated resource sharing significantly increase throughput and model density.

Together, AMD and Rapt enable enterprises to build, optimize, and deploy AI models up to 4× faster, improving overall pipeline efficiency while reducing infrastructure overhead.

Mirantis collabroates with AMD through the k0rdent platform, to deliver GPU-aware Kubernetes orchestration across bare-metal, hybrid, and edge environments.

With full AMD ROCm™ support and integration of the AMD GPU Operator, this collaboration simplifies the deployment and management of AI/HPC workloads—automating driver and stack updates, resource partitioning, and scaling.

Together, they provide enterprises with a streamlined, cloud-native infrastructure for AI/ML orchestration that accelerates DevOps workflows and enhances cluster efficiency on AMD Instinct hardware.

dstack and AMD collaborate to simplify and accelerate AI orchestration on AMD Instinct™ GPUs. Deep AMD ROCm™ integration and AMD hardware support enable dstack’s AI-native control plane to manage containerized workloads—eliminating much of the complexity of Kubernetes or Slurm.

With support for ML experiments, training, and inference, dstack offers unified orchestration across cloud and on-prem environments, provisioning and releasing infrastructure resources efficiently.

The collaboration gives developers plug-and-play access to scalable AI deployments on AMD Instinct GPUs, cutting operational friction and cost while maintaining flexibility and portability.

UbiOps and AMD collaborate to bring high-performance, scalable AI inference to production using AMD Instinct™ GPUs. As a powerful deployment and serving platform, UbiOps leverages AMD Instinct GPUs to handle large-scale, compute-intensive workloads with exceptional speed and efficiency.

Through this integration, users benefit from fast model execution, reduced infrastructure costs, and robust performance—even for the most demanding inference scenarios.

This collaboration empowers developers and customers to deploy cutting-edge AI solutions at scale with confidence and reliability.

AMD and Saturn Cloud are working to make AMD Instinct™ GPU capacity easy to consume across clouds by combining the open AMD ROCm™ software stack for AI and HPC with Saturn Cloud’s code-first MLOps platform.

AMD supplies the AMD Instinct GPUs and AMD ROCm stack—drivers, tools, libraries, and framework integrations—while Saturn Cloud provides both a unified hosted SaaS service, deployed in Saturn Cloud’s own environment, and Saturn Cloud Enterprise, a single-tenant deployment installed directly into each customer’s AWS, Azure, GCP, Oracle, Nebius, Crusoe or other supported cloud account.

Teams can quickly prototype AMD ROCm based workloads on the hosted service, then move into a dedicated Enterprise installation for production so that data, networking, and GPU resources remain within their own cloud boundary while they continue using the same notebooks, jobs, and deployment workflows.

This joint model lowers the operational barrier to adopting AMD Instinct GPUs and AMD ROCm at scale and gives the industry a clear, governed path from evaluation to production on an open, multi-vendor AI infrastructure.

SGLang and AMD collaborate to accelerate next-generation language model development and inference on AMD Instinct™ GPUs. Optimized for high-performance LLM interaction and reasoning, SGLang runs efficiently on AMD hardware through deep integration with the AMD ROCm™ software stack.

This partnership delivers low-latency, high-throughput performance for advanced use cases such as code generation, multi-turn dialogue, retrieval-augmented generation (RAG) and much more, empowering developers with an open, scalable solution to experiment, iterate, and deploy powerful language models across AI research and customer applications.

Embedded LLM and AMD collaborate through the TokenVisor monetization platform, specifically designed for AMD Instinct™ GPUs within the neocloud ecosystem. Launched at AMD's Advancing AI 2025 conference, TokenVisor enables GPU providers to deploy, manage, and monetize LLM services with custom pricing, real-time monitoring, and automated billing capabilities.

This collaboration accelerates time-to-revenue for AMD AI infrastructure investments while providing enterprise-grade management tools that leverage AMD ROCm™ frameworks for efficient workload orchestration and resource allocation.

Seekr and AMD collaborate through Oracle Cloud Infrastructure to deliver trusted AI solutions using AMD Instinct™ MI300X GPUs for enterprise and government applications.

Through the SeekrFlow™ platform and AMD ROCm™ optimization, this collaboration enables secure, transparent agentic AI deployment with advanced explainability capabilities for critical sectors including satellite services and national security applications.

The collaboration provides organizations with cost-effective alternatives for processing decades of imagery and sensor data while maintaining data sovereignty and security standards required for government and enterprise use cases.

Through AMD hardware acceleration and AMD ROCm frameworks, Seekr delivers next-generation vision-language foundation models optimized for edge deployments and multi-node training scenarios.

LlamaIndex collaborates with AMD through AMD ROCm™ integration for building retrieval-augmented generation (RAG) applications on AMD Instinct™ GPUs and AMD Radeon™ graphics.

This collaboration enables developers to construct comprehensive AI pipelines from document ingestion to query engines using LlamaIndex framework with AMD hardware acceleration. The collaboration provides researchers and developers with tools for building knowledge-augmented AI applications including chatbots, question-answering systems, and AI agents while leveraging AMD cost-effective GPU computing capabilities.

Through AMD ROCm compatibility with popular inference engines like Ollama, LlamaIndex users access enhanced performance for RAG workflows across diverse AMD computing environments including enterprise deployments and edge applications.

Canonical and AMD collaborate to deliver a secure, enterprise-grade AI platform on Ubuntu with AMD Instinct™ GPUs. Through AMD ROCm™ software integration and optimized driver support, this collaboration enables seamless GPU acceleration for AI workloads across Linux environments.

Canonical’s containerized infrastructure and AMD’s high-performance GPUs provide IT teams with a stable, scalable foundation for deploying and managing AI applications—combining performance, reliability, and security in production environments.

Red Hat and AMD collaborate to accelerate enterprise AI innovation by integrating AMD Instinct™ GPU support into Red Hat’s hybrid cloud and Kubernetes platforms.

With full enablement of AMD Instinct GPUs in Red Hat OpenShift AI and out-of-the-box support for Red Hat’s new AI Inference Server, this partnership delivers scalable, high-performance AI model deployment across on-premises and multi-cloud environments.

By combining AMD’s cutting-edge GPU hardware with Red Hat’s enterprise-grade software stack, the collaboration empowers organizations with greater flexibility, performance, and choice in deploying advanced AI inference workloads.

SUSE and AMD collaborate to create comprehensive enterprise Linux and Kubernetes support for AMD Instinct™ GPUs and AMD ROCm™. This collaboration provides organizations with stable open-source, enterprise-grade platforms for AI and HPC workloads through native package management, official repository support, and integrated container orchestration via SUSE AI, an open enterprise AI platform that is built on Rancher Prime.

The collaboration enables enterprises to deploy AI applications within SUSE-based infrastructure while leveraging AMD open-source ecosystem for cost-effective GPU computing. Through AMD ROCm integration with SUSE Linux Enterprise Server (SLES), organizations access enhanced capabilities for AI/ML development, high-performance computing, and cloud-native workload deployment across data center and edge environments.

VMware by Broadcom and AMD collaborate to advance AI for enterprises with a private, secure, and high-performance AI infrastructure. We are working continuously on improving how VMware Cloud Foundation (VCF) works together with AMD Enterprise AI software and AMD Instinct™ GPUs.

Together this will enable privacy and security of infrastructure, simplify infrastructure management and streamline AI model deployment for enterprises.

Docker and AMD collaborate through container support for AMD ROCm™ software and AMD Instinct™ GPU acceleration within containerized environments.

This collaboration enables developers to build, deploy, and scale AI applications using Docker containers while leveraging the power of AMD Instinct GPUs. It provides standardized containerization for AMD AI workloads, enabling portable and scalable deployment across diverse computing environments.

Kubernetes ecosystem partners with AMD through AMD ROCm™ integration and AMD Instinct™ GPU support via device plugins and operators.

This collaboration enables container orchestration for AI workloads that leverage AMD accelerators within Kubernetes clusters.

The partnership provides organizations with cloud-native deployment capabilities for AMD-powered AI applications, enabling scalable and manageable AI infrastructure deployment.

CIQ and AMD collaborate to deliver an AMD-optimized Rocky Linux distribution with pre-validated AMD Instinct™ GPU drivers and AMD ROCm™ software pre-integrated, enabling enterprises and HPC centers to deploy AMD-powered AI and high-performance computing infrastructure from day zero without custom integration work.

This collaboration spans CIQ’s full open-source infrastructure stack—including Rocky Linux for the OS, Warewulf Pro for cluster provisioning, Fuzzball for workload orchestration, and Apptainer for containerization—each layer validated and supported for AMD Instinct GPU deployments.

Rocky Linux is an officially supported operating system in AMD ROCm compatibility matrix for AMD Instinct MI300X and MI300A GPUs, and this partnership deepens that foundation with enterprise support, lifecycle management, and a reproducible open-source alternative to proprietary Linux stacks.

DriveNets and AMD deliver a validated full-stack AI Networking reference architecture that offers a maximum-performance, open alternative to NVIDIA's vertical stack. 

AMD Instinct MI355X and MI455X GPUs and AMD Pensando Pollara NICs, paired with DriveNets' scheduled Ethernet Fabric, deliver a fully integrated networking stack optimized for large-scale AI training and inference. 

DriveNets' custom RCCL optimizations for AMD Instinct reduce collective communication latency, driving higher GPU utilization and lower cost-per-token across distributed workloads.

As an open, standards-based Ethernet alternative to proprietary tooling, the joint solution gives cluster builders the performance of a vertically integrated stack — with lower TCO, no vendor lock-in, and ready to deploy in production today. 

Ansys, now part of Synopsys, collaborates with AMD to optimize Ansys Fluent® computational fluid dynamics and Ansys Mechanical™ finite element analysis software for AMD Instinct™ GPUs and AMD ROCm™ platforms.

This partnership enables engineering simulation workloads to leverage AMD GPU acceleration for faster solver performance and larger model capabilities.

The collaboration provides engineers with cost-effective alternatives for compute-intensive simulation workflows while maintaining accuracy and reliability standards.

Cholla and AMD collaborate to advance GPU-accelerated astrophysical simulation on AMD Instinct™ GPUs. Originally developed for large-scale galaxy formation studies, Cholla—a massively parallel, GPU-native hydrodynamics code—was successfully ported from CUDA to the AMD ROCm™ HIP platform for AMD Instinct GPUs.

This transition achieved strong performance on AMD Instinct GPUs even without extensive optimization, highlighting the efficiency of the HIP translation layer and its scalability for next-generation AMD Instinct hardware.

By leveraging ROCm’s open software ecosystem and the compute power of AMD Instinct GPUs, researchers can run high-resolution cosmological simulations on exascale systems like Frontier, accelerating discoveries in astrophysics and enabling open, high-performance scientific computing.

Amber and AMD collaborate to accelerate biomolecular simulations on AMD Instinct™ GPUs through AMD ROCm™ software integration. As a leading molecular dynamics suite for computational chemistry and drug discovery, Amber has been optimized to run efficiently on AMD ROCm software platform—delivering GPU-accelerated performance for molecular dynamics, Quantum Mechanics/Molecular Mechanics, and constant pH simulations.

With the power of AMD Instinct GPUs, researchers can execute large-scale, high-precision simulations with improved speed and scalability.

This collaboration provides the life sciences community with open, cost-effective, and high-performance computing alternatives for drug design, protein folding, and biomolecular research.

Chroma and AMD collaborate to advance GPU-accelerated lattice quantum chromodynamics (QCD) simulations on AMD Instinct™ GPUs. As part of the United States Quantum Chromodynamics Project software stack, Chroma—a high-performance, GPU-native hydrodynamics simulation code—has been containerized and optimized under AMD’s Infinity Hub initiative with full AMD ROCm™ software support for AMD Instinct GPUs.

This collaboration enables theoretical physicists to perform large-scale quantum chromodynamics simulations—from single-node workstations to multi-node clusters—using cost-effective AMD hardware while maintaining compatibility with established research workflows.

By leveraging the open AMD ROCm software ecosystem and the power of AMD Instinct GPUs, researchers can achieve production-ready lattice field theory calculations with higher resolution, scalability, and throughput.

CP2K and AMD collaborate to accelerate quantum chemistry and materials science simulations on AMD Instinct™ GPUs. Through full AMD ROCm™ integration and optimized builds using the AMD Optimizing C/C++ Compiler (AOCC), this collaboration enables GPU-accelerated density functional theory (DFT) calculations, molecular dynamics, and solid-state modeling on AMD hardware.

Available as optimized containers on the AMD Infinity Hub, CP2K offers streamlined deployment for high-performance computing (HPC) workloads. By leveraging AMD ROCm libraries such as hipFFT and hipBLAS, and tuning for multi-node parallelism, researchers achieve substantial speedups over CPU-only runs.

This collaboration empowers computational chemists and materials scientists with scalable, cost-effective tools for studying periodic systems, crystal structures, and complex chemical reactions across diverse research environments.

The MIMD Lattice Computation (MILC) and AMD collaborate to accelerate lattice quantum chromodynamics (QCD) research on AMD Instinct™ GPUs. Supported by the open-source AMD ROCm™ software stack, MILC’s parallel simulation code has been adapted to leverage AMD’s GPU acceleration for large-scale Monte Carlo calculations in particle and nuclear physics.

By utilizing AMD Instinct GPUs, researchers achieve fast and efficient lattice gauge theory computations while preserving the precision required for QCD studies.

This collaboration provides physicists with scalable, cost-effective access to high-performance computing resources—advancing fundamental research into the behavior of quarks, gluons, and the structure of matter.

GROMACS and AMD collaborate to accelerate molecular dynamics simulations on AMD Instinct™ GPUs through full AMD ROCm™ integration and SYCL backend optimization.

This collaboration enables researchers to perform high-performance simulations of proteins, nucleic acids, and lipid membranes on AMD hardware, achieving performance comparable to native HIP implementations. Leveraging the AdaptiveCpp (formerly hipSYCL) runtime, LLVM compiler, and AMD ROCm libraries such as hipFFT and hipBLAS, GROMACS efficiently scales across single-node and multi-GPU cluster environments—including systems like the LUMI supercomputer powered by AMD Instinct GPUs.

Through AMD Infinity Hub containers and AMD ROCm open software ecosystem, researchers gain a streamlined, cost-effective solution for large-scale biomolecular modeling, drug discovery, and biophysical research.

LAMMPS and AMD collaborate to accelerate large-scale molecular dynamics simulations on AMD Instinct™ GPUs through the AMD ROCm™ platform. The Large-scale Atomic/Molecular Massively Parallel Simulator (LAMMPS) leverages AMD GPU acceleration via the Kokkos and GPU packages, both enabled through the HIP programming model.

This integration allows researchers to simulate materials science, soft matter physics, and biomolecular systems with exceptional scalability and performance. With ROCm optimization and the power of AMD Instinct GPUs, LAMMPS achieves efficient parallel execution across single-node and multi-node systems, including exascale machines like Frontier.

Through AMD Infinity Hub containers and AMD ROCm developer ecosystem, scientists gain flexible, cost-effective tools for exploring atomic-scale phenomena, mechanical properties, and phase transitions in complex materials.

NAMD and AMD collaborate to accelerate large-scale biomolecular simulations on AMD Instinct™ GPUs through full integration with the AMD ROCm™ platform.

Developed by the Theoretical and Computational Biophysics Group at the University of Illinois, NAMD is a highly scalable molecular dynamics program used for simulating large biological systems such as proteins, membranes, and viruses.

Through ROCm and the HIP framework, NAMD has been optimized to run efficiently on AMD Instinct GPUs, achieving significant performance and energy-efficiency gains. By leveraging the power of AMD Instinct GPUs, NAMD supports both GPU-offload and GPU-resident modes for optimal performance across diverse simulation scales.

This collaboration empowers researchers in academia and industry with cost-effective, high-performance tools for studying protein folding, drug interactions, and complex biomolecular dynamics.

SPECFEM3D Cartesian and AMD collaborate to accelerate large-scale seismic simulations on AMD Instinct™ GPUs through full integration with the AMD ROCm™ software platform.

This collaboration adds GPU acceleration to SPECFEM3D using the HIP programming framework, enabling efficient deployment across AMD Instinct GPUs. By offloading seismic wave propagation and earthquake modeling calculations to GPUs, researchers achieve significant performance gains for full-waveform inversion and high-resolution 3D seismic imaging.

Through AMD ROCm optimized libraries and multi-GPU scalability, geophysicists can conduct advanced studies in earthquake mechanics, seismic hazard analysis, and subsurface exploration on cost-effective, high-performance AMD hardware—scaling seamlessly from local clusters to exascale systems like Frontier and LUMI.

OpenMM and AMD collaborate to accelerate molecular dynamics simulations on AMD Instinct™ GPUs through full integration with the open-source AMD ROCm™ software platform.

By leveraging the HIP framework, OpenMM delivers native GPU acceleration on AMD Instinct hardware for high-performance biomolecular dynamics and scientific modeling.

This collaboration enables researchers to perform custom simulations with flexible force field implementations and advanced algorithms for drug discovery, protein engineering, and biophysical research.

With AMD ROCm optimized libraries and support for AMD Instinct GPUs, OpenMM provides an open, scalable, and cost-effective platform for large-scale molecular modeling across academic, government, and pharmaceutical research environments.

Stone Ridge Technology and AMD collaborate to accelerate high-performance reservoir simulation on AMD Instinct™ GPUs through the AMD ROCm™ software platform. Stone Ridge Technology’s flagship ECHELON software has been enabled to provide GPU-native performance for AMD Instinct GPUs.

This collaboration provides the energy industry with advanced, scalable simulation capabilities for complex subsurface modeling, including hydrocarbon recovery, carbon capture, and hydrogen storage.

By leveraging AMD ROCm developer tools, optimized math libraries, and high-bandwidth AMD Instinct GPUs architecture, ECHELON delivers exceptional performance and cost efficiency on modern high-performance computing (HPC) systems—empowering customers with open, flexible alternatives for large-scale scientific and engineering workloads.

OpenFold and AMD collaborate to accelerate open, scalable protein structure prediction through high-performance computing. Optimized for AMD Instinct™ GPUs and the open-source AMD ROCm™ software stack, OpenFold enables researchers to train and deploy deep learning models for structural biology with exceptional efficiency.

This collaboration powers large-scale protein modeling pipelines such as ElMerFold on supercomputers like Frontier and El Capitan, achieving unprecedented throughput for structure prediction and bioinformatics research.

By combining the AMD open software ecosystem with energy-efficient GPU acceleration, OpenFold and AMD advance open science, enabling faster, more accessible discoveries in drug development and molecular biology

QMCPACK and AMD collaborate to accelerate quantum Monte Carlo simulations on AMD Instinct™ GPUs through the AMD ROCm™ software platform. QMCPACK, a leading open-source suite for electronic structure and many-body calculations, has been optimized to take advantage of AMD Instinct GPUs with enhanced kernels, advanced memory management, and scalable multi-GPU execution.

By leveraging AMD ROCm compilers, math libraries, and high-bandwidth GPU memory, researchers achieve substantial performance gains for quantum chemistry and materials science workloads.

This collaboration enables computational physicists and chemists to perform large-scale quantum mechanical simulations—from individual workstations to world-class supercomputers—using the open, high-performance computing ecosystem from AMD.

Devito and AMD collaborate to accelerate seismic imaging and computational geophysics on AMD Instinct™ GPUs through full integration with the AMD ROCm™ software platform.

Built as a high-level Python framework for finite difference computations, Devito automatically generates optimized parallel code using backends such as HIP, MPI, and OpenMP to deliver state-of-the-art performance on AMD hardware.

This collaboration enables geoscientists to perform high-resolution seismic modeling, full waveform inversion, and subsurface imaging efficiently on AMD Instinct GPUs and AMD EPYC™ server CPUs. By leveraging AMD ROCm open software ecosystem, Devito combines developer productivity with scalable GPU acceleration—empowering researchers to advance exploration, energy, and geophysical research workflows.

Siemens collaborates with AMD to optimize Simcenter™ STAR-CCM+™ multiphysics computational fluid dynamics software for AMD Instinct™ GPUs using AMD ROCm™ acceleration.

This collaboration enables industrial CFD simulations to achieve faster solve times and handle larger, more complex models through AMD GPU acceleration. The collaboration provides engineering teams with enhanced simulation capabilities while offering cost-effective alternatives to traditional GPU computing approaches.

Raja performance portability framework collaborates with AMD through AMD ROCm™ integration enabling applications to leverage AMD Instinct™ GPUs alongside diverse computing architectures.

This collaboration provides developers with unified programming models that maintain code portability while accessing AMD accelerator capabilities for high-performance computing applications. The collaboration enables efficient GPU computing development through the AMD open-source ecosystem while supporting legacy codebases and future hardware evolution.

Through AMD ROCm programming interfaces, Raja users access consistent performance across heterogeneous computing environments.

Trilinos scientific computing framework and AMD collaborate through ROCm™ support for AMD Instinct™ GPU acceleration of large-scale scientific applications.

This collaboration enables researchers to leverage AMD GPUs for solving complex linear algebra problems, optimization challenges, and multiphysics simulations using established numerical algorithms. The collaboration provides computational scientists with robust tools for engineering and scientific computing while maintaining performance portability across diverse hardware platforms.

Through AMD ROCm mathematical libraries, Trilinos users access enhanced computational capabilities for demanding scientific applications.

PyFR computational fluid dynamics framework collaborates with AMD through AMD ROCm™ compatibility for AMD Instinct™ GPU acceleration of high-order finite element simulations.

This collaboration enables researchers to perform efficient CFD calculations for aerospace, automotive, and scientific applications using AMD accelerators with Python-based workflow integration. The collaboration provides computational fluid dynamicists with modern tools for solving complex flow problems while maintaining flexibility for custom simulation requirements.

Through AMD optimization libraries, PyFR users access enhanced performance for demanding fluid dynamics calculations.

Modern Fortran compiler ecosystems and AMD collaborate to bring GPU acceleration to HPC applications through the AMD ROCm™ software platform and AMD Instinct™ GPUs. With the introduction of the AMD ROCm Fortran compiler and HIP Fortran modules, scientists and engineers can now accelerate both legacy and modern Fortran codebases on AMD GPUs using standard programming models such as OpenMP for GPU offloading.

This collaboration enables researchers to run long-standing scientific and engineering codes efficiently on AMD Instinct GPUs while maintaining full compatibility with existing workflows.

By leveraging AMD ROCm compiler infrastructure, unified shared memory, and optimized math libraries, Fortran developers gain a powerful, portable, and forward-looking path to GPU acceleration—bridging decades of scientific software innovation with next-generation AMD hardware.

The Grid lattice quantum chromodynamics (QCD) framework and AMD collaborate to advance high-performance theoretical physics simulations on AMD Instinct™ GPUs through full integration with the AMD ROCm™ software platform.

Written in modern C++, Grid leverages HIP to deliver performance portability across diverse GPU architectures, enabling researchers to run large-scale lattice gauge theory and quantum field theory calculations efficiently on AMD hardware.

The collaboration provides theoretical physicists with scalable, data-parallel tools for studying the interactions of quarks and gluons, using optimized GPU routines and memory layouts for demanding lattice computations. Through AMD Instinct GPUs and the AMD ROCm ecosystem, Grid achieves strong performance and flexibility from single-node workstations to exascale systems—empowering next-generation research in particle and nuclear physics.