One of AMD's biggest and underrated announcements at Computex was ROCm support for Strix Halo and RDNA 4 GPUs, which could change the way we use AI.
AMD's ROCm Software Stack Now Focused On Consumer-Grade Products, RDNA 4 To Benefit Tremendously With The New Update
When it comes to AI at a client level, there's no doubt that AMD has done some serious work, not just with its hardware capabilities, but also by creating an ecosystem for consumers. There were doubts about whether ROCm would be supported on Team Red's RDNA 4 GPUs in the past, but it seems like Team Red decided to drop us a surprise by announcing their new ROCm 6.4.1, which not only supports RX 9070 series GPUs, but also Strix Halo APUs, making the software stack one of the most capable ones out there at a consumer level.
AMD's ROCm is positioning itself to be a viable alternative to NVIDIA's CUDA, and in terms of how it can benefit the users of the firm's Strix Halo APUs, it is reported that by bringing support for multiple AI libraries for consumer-grade products, AMD will allow them to leverage on the onboard XDNA 2 AI engine. And with ROCm, AI performance would be unlocked for 40 RDNA 3.5 CUs and 16 Zen 5 cores with AVX512, marking a significant performance improvement, especially in inference and training applications.
As for RDNA 4 GPUs, the ROCm will now enable them to utilize the onboard Compute Units and AI accelerators for several applications. Along with it, RX 9000 series GPUs can now run frameworks like PyTorch 2.5/2.6 and Megatron-LM, allowing the average consumer to run LLMs and even Stable Diffusion locally, marking a major pivot from AI for professionals to AI for all. This could start an edge AI revolution if AMD manages to nail future development of the ROCm software stack, and for users, well, they are in for a treat.
Interestingly, Microsoft has just made WSL (Windows Subsystem for Linux) open-sourced, and it is claimed that the new ROCm version works pretty decently with a WSL environment, which opens up even more opportunities. WSL support enables seamless integration of ROCm’s AI tools, which are Linux-focused, into Windows, which will allow developers to leverage a familiar environment to access and develop the ROCm ecosystem, fueling more adoption moving into the future.
AMD finally announced that they are also expanding ROCm's support across Linux distributions such as OpenSuSE and will open it up to other options such as Ubuntu by H2 2025. This is surely a beginning to AMD's pursuit of rivaling NVIDIA's CUDA ecosystem. The focus here is more on the consumer segment since Team Red has a robust hardware portfolio, which works out perfectly for them.
About the author: Muhammad Zuhair is a hardware and technology reporter for Wccftech, specializing in the semiconductor industry and the complex interplay between technology, manufacturing, and geopolitics. His coverage focuses on the corporate strategies and technological roadmaps of industry giants like TSMC, NVIDIA, Samsung, and Intel. Zuhair's expertise lies in deconstructing complex topics such as fabrication nodes (e.g., 2nm process), the economic impact of policies like the CHIPS Act, and the strategic development of AI infrastructure from NVIDIA, AMD and Intel.
Follow Wccftech on Google to get more of our news coverage in your feeds.
