Google Tensor G3 made its debut with the launch of the Pixel 8 (curr. $549 on Amazon) and Pixel 8 Pro (curr. $799 on Amazon), and it was a solid upgrade over its predecessors. While it doesn’t meet the performance scores of other Android and iPhone flagship SoCs, such as the Snapdragon 8 Gen 3, MediaTek Dimensity 9300, and Apple A17 Pro, it does have the lead in some aspects.
For example, the Tensor G3 is now confirmed to have hardware-accelerated AV1 encoding support, a first for smartphone SoCs. To compare, most of the Android smartphones default to H.264, while some flagships come with H.265 video encoding support.
Compared to H.264 and H.265, AV1 encoding leads to improved data compression. To be specific, according to the tests done by Facebook (Meta), the data compression rate on AV1 is up to 50% better. For those wondering, better data compression can reduce the size of the files, making high-resolution video captures such as 4K 60 FPS recordings more storage-efficient on smartphones.
This revelation of Google Pixel G3 SoC comes from Mishaal Rahman, an Android expert, who dug into code snippets to find hardware-accelerated AV1 encoding support. It follows the leaks we had before the launch of the chipset, which suggested that the SoC hardware-level support for AV1 encoding at 4K 30 FPS.
Mishaal points out that 4K 30 FPS support was true at one point, but Google updated the encoder to target 4K 60 FPS right before the launch of the Pixel 8 series. However, even with hardware-accelerated AV1 encoding support, Google devices with Tensor G3 SoC don’t take advantage of it. It’s mostly due to compatibility concerns, as H.264 is still the most widely-used encoder.
The Tensor G3 is the first smartphone SoC to support hardware-accelerated AV1 encoding. It supports AV1 encoding at up to 4K60. No application (including the Pixel Camera app) takes advantage of this, though, likely due to a lack of platform support.
— Mishaal Rahman (@MishaalRahman) February 23, 2024
I know this technically… pic.twitter.com/T2EiSCYLKP
Source(s)
Mishaal Rahman (tweet embedded above) via: IT Home (machine translated from Chinese)