AMD expands AI and ONNX support to more Radeon GPUs

Radeon RX 7900 XTX
(Image credit: AMD)

AMD's has released ROCm 6.0.2, its open-source stack platform, along with Linux driver 23.40.2 expanding support for desktop client GPUs and adding ONNX Runtime support. This update broadens the accessibility of artificial intelligence (AI) development tools to a wider range of users and improves the flexibility of AI model development and inference.

Starting from the ROCm 5.7, AMD's software platform for GPU-accelerated computing already gained support for the Radeon RX 7900 XT, Radeon RX 7900 XTX, and Radeon Pro W7900 boards, which are among the best graphics cards available today. Now, ROCm 6.0.2 expands this support to the Radeon Pro W7800 and the Radeon RX 7900 GRE. This expands the number of developers who could use AMD's ROCm 6.0 platform for their AI development efforts on their desktops. The Linux driver is compatible with 64-bit Ubuntu 22.04.3 HWE.

Another notable addition in the ROCm 6.0 platform is support for ONNX Runtime (Open Neural Network Exchange Runtime). This feature enables conversion of AI models between various machine learning frameworks, which greatly enhances the interoperability and flexibility of AI development. As a result, users can now perform inference tasks on a wider range of source data directly on local AMD Radeon hardware, streamlining the workflow for AI model deployment.

The update also adds INT8 data type to MIGraphX, AMD's proprietary graph inference engine, expanding the range of supported data types (which currently includes FP32 and FP16) to enhance flexibility and performance. The inclusion of INT8 support is particularly beneficial for applications that require lower precision but higher throughput.

Finally, AMD's ROCm 6.0 brings mixed precision capabilities with FP32/FP16 to PyTorch machine learning training workflows, which enables faster model development and iteration.

"These are exciting times for anyone deciding to start working on AI," wrote David Diederichs, Product Marketing Manager of Pro Software at AMD, in a blog post. "ROCm for AMD Radeon desktop GPUs is a great solution for AI engineers, ML researchers and enthusiasts alike and no longer remains exclusive to those with large budgets. AMD is determined to keep broadening hardware support and adding more capabilities to our Machine Learning Development solution stack over time."

Anton Shilov
Freelance News Writer

Anton Shilov is a Freelance News Writer at Tom’s Hardware US. Over the past couple of decades, he has covered everything from CPUs and GPUs to supercomputers and from modern process technologies and latest fab tools to high-tech industry trends.

  • gg83
    By the headline, I thought amd expanded support to more than the 7900gre and a pro gpu. Why wouldn't the pro gpu already support ROCm? Is it older?
  • bit_user
    gg83 said:
    By the headline, I thought amd expanded support to more than the 7900gre and a pro gpu. Why wouldn't the pro gpu already support ROCm? Is it older?
    Perhaps you're confusing ONNX support with ROCm? ONNX translation is a purely software feature that has no intrinsic relation to any particular hardware spec. Hence, they can allow ONNX translation for all supported GPU models in one go. Yes, it's weird to stipulate that in the headline, I suppose.