AMD’s ongoing efforts to enhance its ROCm software stack have sparked renewed interest among users and developers, particularly regarding the potential expansion of support from Linux to Windows. For years, the community has been eagerly anticipating AMD’s commitment to this initiative, especially after the company initially hinted at bringing ROCm support to Windows. Recently, AMD’s Vice President of AI Software, Anush Elangovan, confirmed that the company is indeed looking to broaden ROCm support for additional GPUs on the Windows platform.
AMD’s Commitment to Expanding ROCm Support
Currently, ROCm support on Windows is limited to specific models, including select AMD Instinct GPUs and a few Radeon GPUs, such as the RX 7900 XT and XTX. Unfortunately, this means that many users with Radeon GPUs are unable to take advantage of ROCm’s capabilities, particularly those with the newer RX 9000 series, which remain unsupported. As of now, the latest version available for Windows is 6.2.4, but the compatibility issues persist.
Elangovan’s affirmative response to the inquiry about extending ROCm support suggests a promising shift in AMD’s strategy. While he did not provide extensive details, his acknowledgment indicates a willingness to enhance the software’s reach across a broader range of Radeon GPUs. In contrast, users with RDNA 2 GPUs on Linux have been able to utilize ROCm without the same limitations.
Despite the potential for improved support, users have reported various challenges when using ROCm on Windows. Issues such as crashes, driver timeouts, and application freezes can hinder the experience, particularly for those using GPUs that fall below the RX 7900 GRE threshold. This situation leaves many users unable to fully leverage the powerful tools ROCm offers.
As AMD continues to address these complexities, the prospect of enhanced support for ROCm on Windows could open new avenues for deep-learning tasks, particularly for owners of older GPUs. The anticipated support for the latest RDNA 4 GPUs also holds significant promise for developers and users operating within the Windows ecosystem.
In a related development, tinygrad, a company focused on AI solutions, has announced that it will receive two MI300X boxes directly from AMD. This collaboration represents a significant step forward for the AI segment, as tinygrad aims to create a fully sovereign AMD stack, enhancing the capabilities of its offerings.
The developer’s perspective highlights a growing sentiment that if AMD can successfully refine its software, it may challenge the current market valuation disparities between itself and NVIDIA. With the demand for AI technologies on the rise, the potential for AMD to carve out a more prominent position in the market appears increasingly viable.
