AMD Ryzen AI Developer Platform OS updated with ROCm 10.0, Linux 7.2 (Phoronix)

Releases·Tool Release

AMD releases ROCm 10.0, its open compute stack aimed at agentic AI workloads

AMD released ROCm 10.0, a major version of its open GPU compute stack, framed around agentic AI workloads. The jump from 7.14 to 10.0 came just a month after the previous release, and llama.cpp support is pending via an open pull request.

AMD released ROCm 10.0, a major version of its open GPU compute stack, framed around agentic AI workloads. The jump from 7.14 to 10.0 came just a month after the previous release, and llama.cpp support is pending via an open pull request.

Releases·Tool Release

AMD tags ROCm 7.14 'TheRock' tech preview for latest GPU compute stack

AMD has tagged a tech preview of ROCm 7.14, codenamed 'TheRock', for its latest GPU compute stack. The preview is available for testing and includes support for newer AMD GPUs. This release aims to improve performance and compatibility for AI and HPC workloads on AMD hardware.

AMD has tagged a tech preview of ROCm 7.14, codenamed 'TheRock', for its latest GPU compute stack. The preview is available for testing and includes support for newer AMD GPUs. This release aims to improve performance and compatibility for AI and HPC workloads on AMD hardware.

Releases·Tool Release

AMD's llama.cpp branch boosts prompt processing up to 2x on Strix Halo

AMD maintains its own llama.cpp branch with ROCm/Hip patches that can more than double prompt processing speed on dense models. A user reports 550 tokens/s with a 14B dense model versus 230 tokens/s on standard llama.cpp, though token generation is about 15% slower than with Vulkan.

AMD maintains its own llama.cpp branch with ROCm/Hip patches that can more than double prompt processing speed on dense models. A user reports 550 tokens/s with a 14B dense model versus 230 tokens/s on standard llama.cpp, though token generation is about 15% slower than with Vulkan.

Releases·Tool Releasefeatured

Unsloth announces official AMD GPU support for local LLM fine-tuning and inference

Unsloth now officially supports AMD hardware for local inference, fine-tuning, reinforcement learning, and deployment. The open-source tool works on Windows, Linux, and WSL with Radeon RX 9000/7000 series, Instinct MI350/MI300 GPUs, Strix Halo/Ryzen AI Max systems, and AMD CPUs for GPU-free inference. It claims up to 70% less VRAM for training and 80% less for reinforcement learning.

Unsloth now officially supports AMD hardware for local inference, fine-tuning, reinforcement learning, and deployment. The open-source tool works on Windows, Linux, and WSL with Radeon RX 9000/7000 series, Instinct MI350/MI300 GPUs, Strix Halo/Ryzen AI Max systems, and AMD CPUs for GPU-free inference. It claims up to 70% less VRAM for training and 80% less for reinforcement learning.

A community member published a complete guide for running PyTorch with ROCm on AMD GPUs, targeting RDNA3 cards like the 7900 XTX. The guide covers installation, VRAM tuning, and troubleshooting to help local AI users get better performance.