Google's Tensor SoC powering the new Pixel 6 line brings ML capabilities to the handsets and more
We may earn a commission if you make a purchase from the links on this page.
Google introduced its homegrown Tensor SoC today and the chip will power both the Pixel 6 and Pixel 6 Pro. The company noted that it hadn't been able to bring its Machine Learning capabilities with the chips it had used on past Pixel models (the Qualcomm Snapdragon line). But by designing its own chips and optimizing them for Google's Machine Learning, the new Pixels will have features that Google could not offer on the phone before.
For example, Google's Machine Learning Engine, TPU, is included in Tensor. This engine was designed by Google Research for Google Research. And the decisions made in building the CPU and GPU were done with ML in mind so that the Pixel 6 line can deliver advanced computational photography.
For those unfamiliar with Machine Learning, it is the ability of computer algorithms to improve over time through experience and data. Machine Learning is considered part of the world of Artificial Intelligence (AI). Tensor improves the quality of speech translation on the new Pixels by 18%.
Pre-order your new Pixel 6 or Pixel 6 Pro now and take advantage of the Tensor chip
Google notes that the Tensor chipset helps the Pixel 6 and Pixel 6 Pro provide new experiences "including Motion Mode, Face Unblur, Speech enhancement mode for videos and HDRnet for videos.
The Tensor SoC, unsurprisingly, is much faster than the Pixel 5's Snapdragon 765G
The Tensor's CPU is made of eight cores with two high-performance cores, two mid cores, and four high-efficiency cores. The GPU included with the Tensor CPU sports 20 cores for advanced graphics on the most popular Android games. The chip includes a context hub allowing the Pixel to provide ML and ambient features such as the phone's always-on display and the Now Playing feature (which identifies songs playing in the background) to run without draining the battery.
Thanks to the Tensor chip, Google Assistant uses the most accurate Automatic Speech Recognition (ASR) ever created by Google. And Google will allow features such as Live Caption or Recorder to access ASR without consuming too much battery life.
Google notes that peak CPU and GPU benchmarks may look nice, but they don't always reflect the real-world experience of an AP. According to the company, the CPU and GPU experiences using Tensor are 80% and 370% faster respectively than the experience of the Snapdragon 765G on last year's Pixel 5. Of course, to be fair (and this was not noted by Google), the Snapdragon 765G is not top-of-the-line silicon.
Voice typing is faster than touch typing on the new Pixels thanks to Tensor
The Tensor chip will also help Pixel owners consume 50% less power when the on-device natural language chip is figuring out what task you have just asked Google Assistant to do. And thanks to the Tensor chip, when voice typing, the phone will know that when you say "send," it isn't meant to be part of the message. Thanks to Tensor, voice typing is three times faster than typing on the keyboard with your finger.
The Alphabet unit also confessed that videos filmed using past Pixel models were not up to the quality of its still photos. The company points out that "We’ve always dreamed of getting Pixel video to match the quality of Pixel photos — and Google Tensor has helped us deliver better experiences in each area.
Google states that it built the chip differently than other SoCs. "Google Tensor was built to be a premium system on a chip (SoC) that has everything you would expect from a mobile SoC, and more. So how did we do this? The core experience areas — speech, language, imaging and video — for our new phones are all heterogeneous by nature, meaning they require multiple resources across the entire chip."
The company continued by saying that it "...made sure that Google Tensor was carefully designed to deliver the right level of compute performance, efficiency and security. And with Android 12, we set out to build an OS that lays the foundation for the future of hardware and software working together. You can see this in real-world use cases, like taking amazing videos or understanding more foreign languages."
Things that are NOT allowed: