Intel has published the release of the SVT-AV1 1.0 (Scalable Video Technology AV1) library, which provides an alternative encoder and decoder for the AV1 video encoding format, which uses the hardware parallelization of computations present in modern Intel CPUs. The main objective of the SVT-AV1 is to achieve a level of performance suitable for on-the-fly video transcoding and use in video-on-demand (VOD) services. The code is developed by the OpenVisualCloud project, which also develops the SVT-HEVC and SVT-VP9 encoders, and is distributed under the BSD license.
SVT-AV1 requires at least a fifth generation Intel Core processor (Intel Xeon E5-v4 and newer CPUs). To encode 10-bit AV1 streams in 4K quality, 48 GB of RAM is required, 1080p 16 GB, 720p 8 GB, 480p 4 GB. Due to the complexity of the algorithms used in AV1, encoding this format requires significantly more resources than other formats, which does not allow using the standard AV1 encoder for real-time transcoding. For example, the stock encoder from the AV1 project requires 5721, 5869, and 658 times more computation than the x264 (main profile), x264 (high profile) and libvpx-vp9 encoders.
Among the changes in the new SVT-AV1 release:
- Added support for S-frames (Switching Frame), intermediate frames whose content can be predicted based on previously decoded reference frames from the same video at a higher resolution. S-frames allow you to increase the compression efficiency of live streams.
- Added constant bit rate (CBR) coding control mode for minimal latency.
- Added support for passing color subsampling position information.
- Added the ability to skip denoising images after coarse synthesis.
- Fast decoding support extended to M0-M10 presets.
- Simplified use of the "--fast-decode" option and optimized the first level of fast decoding.
- Improved visual quality of the encoding result.
- Optimized memory consumption.
- Added additional optimizations based on AVX2 instructions.
Source: opennet.ru