Collabora engineers, backed by Valve sponsorship, have successfully ported the open-source Mesa RADV Vulkan graphics driver natively to Microsoft Windows, bypassing Linux subsystem layers and achieving successful gameplay demos in titles like Counter-Strike 2 as detailed in late July 2026 reports.
For years, the graphical divide between Linux and Windows operating systems felt absolute, governed by entirely different underlying architectures and proprietary software stacks. That barrier is beginning to crack in an unexpected way. Software consultancy Collabora detailed Valve’s sponsorship to bring RADV—the community-maintained open-source Mesa Vulkan driver optimized for AMD graphics cards—natively to Microsoft Windows.
Valve and Collabora Target Windows With Open-Source RADV
The project builds directly upon earlier research presented at XDC 2024 by Faith Ekstrand. By moving forward with exploratory work, Valve has been sponsoring Collabora engineers to determine whether the Mesa RADV Vulkan driver can become viable outside of Linux. This open-source driver is widely recognized as a cornerstone of the Linux gaming ecosystem, having played a massive role in the success of AMD-powered portable hardware like the Steam Deck and Steam Machines.
Until now, Windows has had no open-source drivers for any major GPU manufacturer. Bringing RADV across the aisle eliminates the need for translation layers like the Windows Subsystem for Linux (WSL), allowing high-performance graphics code to target Windows directly. Louis-Francis Ratté-Boulianne published a blog post highlighting the initial development phase, demonstrating that the experimental code can already launch Counter-Strike 2 with RADV on Windows.
Navigating the Proprietary Kernel Barrier
Running an open-source User-Mode Driver (UMD) on Microsoft Windows requires deep integration with the operating system’s display driver architecture via WDDM2 integration. However, the engineering team faced a formidable roadblock: porting the Linux amdgpu
kernel driver to Windows was entirely out of scope. Instead, RADV has to interface directly with AMD’s official, proprietary Windows kernel driver.
Because AMD’s official Windows kernel driver is packed with opaque code blocks and lacks public documentation or troubleshooting tools, the Collabora team had to rely heavily on reverse engineering. According to technical breakdowns published on 28 July 2026, engineers built custom utilities—including wddm2-pdd-re
and a specialized WDDM2 logging layer—to inspect command streams and hardware registers.
This meticulous reverse-engineering effort allowed developers to map out undocumented kernel data structures. Yet, relying on private structures remains inherently fragile. As industry reports point out, the user-mode driver and kernel-mode driver ship as a matched pair with zero backward-compatibility guarantees, meaning AMD driver updates can alter these data structures without warning.
Technical Hurdles and Early Feature Support
Despite the architectural friction, the experimental codebase has matured past simple proof-of-concept tests. The development team successfully resolved GPU hangs across multiple hardware generations, addressed MSVC compiler issues such as signed 32-bit enum handling, and confirmed that vendor-specific D3DKMTEscape calls can be safely bypassed in single-GPU setups.
- Sparse resource bindings
- Tessellation and task shaders
- GPU property detection and reporting
- Improved synchronization primitives
While the project has proven concept viability, researchers stress that the work remains in its early stages. Stability improvements and complete feature parity with every specialized GPU capability still require extensive development.
The Long-Term Viability Question for Windows Open-Source Drivers
The ultimate fate of running an open-source Vulkan driver on Microsoft’s dominant operating system hinges on cooperation and long-term maintenance strategies. Because the current setup depends on undocumented data structures that change between official driver releases, developers face a constant maintenance tax.
To achieve true production readiness, software analysts note that RADV will either need AMD to supply a stable, officially documented kernel interface or require developers to maintain a specialized compatibility shim layer sitting permanently between the drivers.
For now, Collabora has made its experimental code available on GitLab for public testing and community contributions, leaving open whether production-level deployment will follow this initial proof of concept.
