Taking on CUDA with ROCm: 'One Step After Another'

eetimes.com

April 12, 2026

6 min read

🔥🔥🔥🔥🔥

60/100

Summary

AMD's ROCm software stack aims to compete with Nvidia's CUDA for data center GPU market share. Success in this endeavor is viewed as a significant challenge due to CUDA's established dominance.

Key Takeaways

AMD's ROCm software stack is crucial for the company's strategy to compete with Nvidia's CUDA in the data center GPU market.
ROCm has undergone significant development over the past two and a half years, evolving from a collection of parts to a unified AI stack called OneROCm.
The integration of OpenAI's Triton framework has facilitated portability between AMD and Nvidia GPUs, reducing the need to convert CUDA code.
AMD is focused on achieving a six-week release cadence for ROCm, aiming for seamless functionality akin to that of Google Chrome.

Read original article

View story →

Coverage & discussions

Hacker News · eetimes.com

Taking on CUDA with ROCm: 'One Step After Another' ↗

Community discussions

Hacker News discussion ↗

Historical count: 5 · 5 saved in sample · Historical sample

Analysis ready

Community Sentiment

Negative

Positives

Efforts to build ROCm with a native musl/mimalloc toolchain are paving the way for high security and privacy workloads, showcasing AMD's commitment to open-source solutions.
Vulkan's backend for Ollama is functioning well, indicating that alternative solutions can effectively support AI workloads despite ROCm's limitations.

Concerns

ROCm's lack of support for older consumer GPUs like the RX 580 limits accessibility, which could hinder its adoption in the AI community.
AMD's ROCm has been criticized for not fully supporting its own graphics cards, leading to instability and bugs that affect AI performance.
The perception that AMD has neglected ROCm for years raises concerns about its competitiveness in the AI space, especially against NVIDIA.