Over a year ago, thanks to the joint efforts of Microsoft, Linaro, and Qualcomm, the ambitious project of porting Blender to the ARM64 architecture running Windows (WoA) began.
With Qualcomm’s significant support as a Patron-level member of the Blender Development Fund, the core development team was able to review and iterate on this project. As a result, Blender can now run on hardware powered by Windows on processors such as the Qualcomm Snapdragon.
This journey began with Blender 4.3, the first release to officially support WoA, and continues today with the addition of Vulkan support for high‑performance viewport rendering in EEVEE. A summary of this journey is available in Anthony Robert’s presentation at Blender Conference 2024.
Vulkan-powered
In order to provide the best viewport and overall rendering performance, the goal is to continue advancing development on the Vulkan graphics API to take full advantage of hardware acceleration components.
Some of the targets that planned as part of this project were:
- Shader optimization to improve UI performance
- Viewport speedup by utilizing tiling, on Adreno GPU architecture
In the longer term (during 2026), the aim is to offer hardware‑accelerated ray tracing for Cycles on Snapdragon GPUs, making use of SYCL.
Benchmarks
With the release of Blender 4.5 LTS and the Vulkan graphics backend, the EEVEE viewport playback and rendering performance has drastically improved.
Playback Speed for the Amy and Tree Creature demo files on Qualcomm® Adreno™ X1-85 GPU 20803
Rendering Speed for the Amy and Tree Creature demo files on Qualcomm® Adreno™ X1-85 GPU 20803
Testing and feedback
If you are using Blender on Windows on Arm, download the latest release from blender.org or get a build from builder.blender.org and give it a try. In order to enable the Vulkan backend, open the Blender Preferences -> System, and select it from the options available. Please report any issues via Blender, using the Help -> Report a Bug menu.
Congratulations to the development team at Linaro and Qualcomm, including Anthony Roberts, Jeffrey Moguillansky, Mark Feldman, Nilesh Shah, and Janardhan Haryadi Ramesh for this contribution. Special thanks to Brecht Van Lommel and Jeroen Bakker for reviewing and coordinating such a large project.
