A superblock is AV1's largest coding unit — a 128×128 pixel block that the encoder can recursively split into smaller pieces all the way down to 4×4. It's the AV1 equivalent of HEVC's ctu (which tops out at 64×64), and the move from 64×64 to 128×128 lets AV1 cover large flat regions of a frame (sky, walls, out-of-focus backgrounds) in a single block with minimal header overhead.
The bigger size matters more than it sounds. A 4K frame contains 510 superblocks of 128×128, versus 2,040 CTUs of 64×64 if it were HEVC. Each block carries some fixed header overhead, so fewer-but-larger blocks save bits when the content is simple. For complex content, the encoder splits the superblock down through a flexible partition tree — supporting not just 4 equal quadrants like HEVC's quadtree, but also rectangular splits (a 128×128 can become two 64×128, or four 32×128, etc.). This non-square flexibility captures the shape of real visual features (a horizon line, a vertical pillar, a wide banner) more naturally than HEVC's purely-square splits.
For a product team, the superblock is invisible plumbing that contributes meaningfully to AV1's ~30 % compression gain over HEVC. There's nothing to configure — the encoder uses superblocks by default. The downside is encoder complexity: evaluating which partitioning to use for each 128×128 region is computationally expensive, which is part of why AV1 encoding (even with SVT-AV1) is several times slower than HEVC encoding. That's the recurring trade-off across codec generations: more block partitioning flexibility produces smaller files, but costs more compute to encode.
