Login to Continue Learning
AMD has initiated work on releasing the ROCm 7 software stack, which was anticipated to challenge NVIDIA’s dominant CUDA ecosystem in the AI industry.
**AMD’s ROCm 7 Sees Addition into GitHub; Launch Expected Soon**
One of the key reasons for NVIDIA’s dominance in AI is its CUDA software suite. This is the primary choice for developers and is exclusive to native hardware, leaving little room for competitors like AMD. However, at the Advancing AI event, AMD announced the next-gen ROCm 7.0, which was said to bring enhanced frameworks and new algorithms to create an alternative to NVIDIA’s CUDA.
According to Phoronix, AMD has added release tags for ROCm 7.0 on GitHub, indicating that a launch is imminent. The tags include `rocm-7.0.0` in ROCm/hip and `rocm-7.0.0` in ROCm/aomp, with HIP, AOMP, and ROCm Libraries all having their 7.0.0 tags ready for display.
**Major Upgrades and Performance Improvements**
The firm claimed that the primary focus of the upgraded compute stack was on inferencing workloads, achieving up to a 3.5x performance uplift in AI workloads compared to ROCm 6. Notably, AMD’s Instinct MI355X outperformed NVIDIA’s Blackwell B200, with a 30% higher FP8 throughput in DeepSeek R1.
The major upgrades include:
– Latest Algorithms & Models
– Advanced Features for Scaling AI
– MI350 series support
– Cluster Management
– Enterprise Capabilities
While there is no defined release date yet, the software stack is expected to launch in the upcoming weeks. The timing will likely depend on how AMD integrates it with its upcoming AI hardware releases.
For more details, refer to [Phoronix](https://www.phoronix.com/news/AMD-ROCm-7.0-Rolling-Out).
