Intel's Open Image Denoise 2.0 Brings SYCL For Xe GPUs, NVIDIA CUDA, AMD HIP

Among Intel's dozens of terrific open-source components -- including the many components making up their oneAPI software suite -- is Open Image Denoise. Open Image Denoise for years has been a terrific, high-performance denoising library for ray-tracing use The software has long been CPU-based while being highly performant thanks to leveraging modern instruction set extensions. Today though Open Image Denoise 2.0 is released and brings GPU acceleration across Intel, AMD, and NVIDIA graphics processors.

Headlining the big Open Image Denoise 2.0 release is GPU acceleration, similar to other Intel oneAPI rendering components that have long been CPU-based but since extended to support GPU acceleration so that modern Intel Xe Graphics / Data Center GPU Flex Series / Intel GPU Max hardware can leverage their capabilities with this software.

Open Image Denoise 2.0 brings SYCL support for use on all Intel Xe DG2/Alchemist GPUs from the consumer desktop Arc Graphics up through the Intel GPU Max series. As with many other Intel oneAPI components, in addition to supporting their own dGPUs they have also enabled support for other graphics vendors too. Complementing the SYCL back-end is OIDn 2.0 also supporting NVIDIA CUDA and AMD HIP for supporting cross-vendor hardware with this denoising library.

The Open Image Denoise 2.0 release also has new API additions, support for asynchronous execution, physical device API for querying supported devices on the system, and a variety of other improvements.

Details on the Open Image Denoise 2.0 release along with downloading the Apache 2.0 licensed source code or the Windows / macOS / Linux binaries can be found via GitHub. If this is your first time learning about this great open-source project, visit OpenImageDenoise.org. I'll be trying out OIDn 2.0 on various CPUs and GPUs for Linux testing as time allows.