Tensor G3 chip AI upgrade: Google Pixel 8 / Pro phone runs a large model locally, increasing the amount of computation by 150x

CTOnews.com, October 4 (Xinhua)-- the new flagship Android phone, the Google Pixel 8 / Pro series, was officially launched at the new product launch this evening, equipped with the Tensor G3 chip and the Titan M2 security chip.

Tensor G3, which has always been associated with AI, runs a more complex ML (machine learning) model, bringing AI enhancements to almost every part of Pixel 8 and 8 Pro, according to Google. It reads web pages aloud for users in a "more natural" voice, and can read them in different languages, and virtual assistants speak more naturally.

Google says Pixel 8's Tensor G3 runs twice as many machine learning models on devices as Pixel 6, the original Tensor chip.

In addition, the Pixel 8 Pro claims to be the first phone to run Google's AI model directly on the device, with 150 times more computation than the largest ML model on Pixel 7.

In addition, the Tensor G3 provides better photo and video capabilities; it can also translate face-to-face conversations, real-time transcribed messages and translated symbols in 49 languages; Google Assistant can now detect and filter out more junk calls; Pixel can detect whether users have been in a serious car accident and call emergency services.

In terms of performance, the Tensor G3 chip is based on Samsung 4nm technology, with 1 × Cortex-X3, 4 × Cortex-A715, 4 × Cortex-A510 CPU cores.

According to the Geekbench 6 score data previously found by CTOnews.com, Google's Pixel 8 Pro phone has a single core score of 1760 and a multi-core score of 4442, which is not a very top level.