SVT-AV1 Default Tile Settings for 4K Video

This article explains the default tile settings applied by the SVT-AV1 encoder when compressing standard 4K resolution video. It details how the encoder configures tile rows and columns by default, the technical reasoning behind these settings, and how they impact both encoding efficiency and parallel processing.

The Default Tile Configuration

By default, the SVT-AV1 encoder sets both --tile-rows and --tile-columns to 0.

In the AV1 video codec standard, tile parameters are expressed in \(\log_2\) units. A setting of 0 translates to \(2^0\), which equals 1. Therefore, for a standard 4K video (3840x2160), the default configuration is:

This means that, out of the box, SVT-AV1 does not split a 4K video frame into multiple tiles; it treats the entire frame as a single, continuous tile.

Why SVT-AV1 Defaults to a Single Tile

Unlike older encoders or other AV1 implementations that rely heavily on tiling to achieve multi-threading, SVT-AV1 utilizes a highly sophisticated internal threading architecture.

1. Advanced Multi-Threading

SVT-AV1 uses segment-based and row-based parallel processing (such as Wavefront Parallel Processing) alongside picture-level parallel processing. This allows the encoder to distribute the workload of a 4K video across multiple CPU threads efficiently without needing to segment the video frame into physical tiles.

2. Compression Efficiency

Tiling introduces boundaries across the video frame. These boundaries restrict in-loop filtering and intra-prediction (the encoder cannot reference pixels across tile borders). By defaulting to a single tile, SVT-AV1 maximizes compression efficiency, resulting in a higher visual quality at a lower bitrate.

When to Override the Default Settings

While a single tile is optimal for encoding efficiency, there are scenarios where you should manually adjust the tile settings for 4K video: