• 2 min read
AMD targets CUDA’s edge with ROCm.AI
AMD’s ROCm.AI uses coding assistants and Hyperloom to optimize AI workloads for Instinct GPUs, with a reported 38% gain on Helios racks.
Image: The Register
AMD has introduced ROCm.AI, a platform designed to help developers use frontier AI models to deploy, debug, and optimize workloads on AMD Instinct GPUs without manually tuning every low-level kernel.
The announcement came at AMD’s Advancing AI event in San Francisco this week, as the company continues to challenge the perception that its GPUs are less capable because they do not run NVIDIA’s CUDA platform. Frameworks such as PyTorch and JAX already let developers write code that runs across hardware, but achieving peak performance still requires work at the CUDA, ROCm, or HIP level.
How ROCm.AI optimizes AMD workloads
ROCm.AI connects existing coding assistants to the tools and documentation needed to optimize AMD hardware. Its automated workload optimization system, Hyperloom, can be invoked with a prompt such as “optimize MiniMax M3 with Hyperloom.”
Recommended reading
3,607 reports show how AI agents misbehave
According to AMD corporate VP of AI software and solutions Anush Elangovan, Hyperloom can:
- Launch an inference server in a Docker container
- Benchmark the workload to establish a baseline
- Profile the application and identify bottlenecks
- Adjust configuration or generate custom CPU kernels
In tests on AMD’s newly launched Helios racks, AMD said the process improved model performance by 38 percent over baseline.
“We want to give you the ability to eke out the maximum performance. This makes it incredibly easy for anyone to consume, debug, profile, and deploy.”
AMD says its machine-readable instruction set architecture specifications help frontier models understand how to program its GPUs. The company is also working with model developers including OpenAI and Anthropic to train systems that natively understand AMD’s programming model.
“We’re not just using the frontier model to generate a kernel. We’re working deeply with frontier model companies so that they natively speak AMD programming.”
ROCm.AI will include a command-line interface and plug-ins for Claude Code, OpenAI Codex, Google’s Antigravity, and Cursor.
AI Editor
Ava covers the rapidly evolving world of artificial intelligence, from foundational models and research labs to the real-world economics of intelligence. With a background in computational linguistics, she cuts through the hype to find out what actually works. She firmly believes that benchmarks are just marketing until reproduced in the wild.
via The Register
