AMD is making significant strides in expanding the capabilities of its ROCm 6.0 software, as it continues to compete with industry giants like NVIDIA and Intel. With updated support for both consumer and professional GPUs, along with the addition of new AI-based features, AMD is ensuring that its software ecosystems remain on par with the competition.
In its latest announcement, AMD is introducing ROCm 6.0 support for two of its RDNA 3 GPUs designed for consumer and professional segments: the Radeon RX 7900 GRE and the Radeon PRO W7800. These GPUs join a range of other supported consumer cards, such as the Radeon PRO W7900, Radeon RX 7900 XTX, and Radeon RX 7900 XT. This expanded support allows AI enthusiasts to explore a wider range of products without being limited to the ultra-enthusiast segment.
Not only does AMD’s ROCm 6.0 software broaden GPU support, but it also adds support for ONNX Runtime, enabling inference to be executed on a broader range of source data available on AMD hardware. The software suite now also supports INT8 compute through MIGraphX, in addition to FP32 and FP16. Furthermore, AMD’s GPUs now have support for CUDA apps through ROCm software via ZLUDA. The company is also continuing its support for PyTorch, further enhancing its FP32 and FP16 mixed precision capabilities.
The progress made by AMD’s software team in strengthening the ROCm 6.0 ecosystem for the AI and ML era is commendable. Despite the competition from established software ecosystems, such as CUDA and OpenAPI, AMD is determined to keep pace and exceed expectations. By making the ROCm 6.0 source code available, AMD is fostering an environment of collaboration and innovation.
Looking ahead, AMD promises to continue expanding the hardware and software capabilities of ROCm 6.0, solidifying its position as a key player in the AI industry. With its commitment to empowering AI enthusiasts with cutting-edge technology, AMD is set to revolutionize the future of AI and machine learning.
AMD’s ROCm 6.0 software: AMD is improving its software, ROCm 6.0, to compete with companies like NVIDIA and Intel. It offers support for both consumer and professional GPUs, with new AI-based features.
RDNA 3 GPUs: AMD introduces support for two RDNA 3 GPUs: Radeon RX 7900 GRE and Radeon PRO W7800. These GPUs join the existing Radeon PRO W7900, Radeon RX 7900 XTX, and Radeon RX 7900 XT, expanding the range of supported consumer cards.
ONNX Runtime: ROCm 6.0 now supports ONNX Runtime, allowing inference to be executed on AMD hardware using a wider range of source data.
INT8 compute: ROCm 6.0 supports INT8 compute through MIGraphX, in addition to FP32 and FP16. This broadens the capabilities of AMD GPUs.
CUDA apps: With ROCm software via ZLUDA, AMD GPUs now have support for CUDA apps, enabling compatibility with a popular programming framework.
PyTorch support: AMD continues its support for PyTorch, enhancing its FP32 and FP16 mixed precision capabilities.
Collaboration and innovation: By making the ROCm 6.0 source code available, AMD encourages collaboration and innovation in the AI and ML community.
Future expansion: AMD plans to further expand the hardware and software capabilities of ROCm 6.0, solidifying its position in the AI industry.
For more information on AMD ROCm 6.0, you can visit their official website: AMD ROCm Software Platform.
The source of the article is from the blog trebujena.net