Google's Pixel 6 and Pixel 6 Pro are now official. The Mountain View corporation confirmed the new smartphones on Monday, confirming part of the previous leaks and leaving several other news. One of the most relevant points is the implementation of its own chip called Tensor , the first SoC designed by Californians.
Of course, Google has not yet given details regarding Tensor's specifications or its performance, but it has opened the door to a very interesting path. The company promises that its fledgling processor will be the key to a new level of harnessing artificial intelligence and machine learning.
This was made known by Rick Osterloh, Google's hardware chief, in a demonstration in front of a handful of US media last week. The manager did not refer to benchmarks or specificities about Tensor's performance, but assured The Verge that it will be “a leader in the market.” However, emphasized that AI will be the point of differentiation with the competition .
“Artificial intelligence is the future of our innovation work, but the problem is that we have run into computational limitations that prevented us from fully fulfilling our mission. So we set out to build a technology platform designed for mobile devices that allowed us to bring the More innovative AI and machine learning to our Pixel users. We set out to make our own SoC to power the Pixel 6. And now, years later, it's almost here, “Osterloh himself posted on Google's blog.
What to expect from Tensor, the Google chip for the new Pixel 6 and Pixel 6 Pro?
According to Sundar Pichai, CEO of Google, the company spent four years developing the new SoC. “Tensor builds on our two decades of computing experience and is our biggest Pixel innovation to date,” he posted on Twitter. Those from Mountain View blindly trust that their new chip is a game changer, and the public is also excited that it is, but there is still a lot of information to be learned.
Osterloh explained to Engadget that Tensor is an ARM chip designed around a Tensor Processing Unit , or TPU, based on the larger versions that Google already implements in its data centers. The name of the SoC is not a coincidence either; it is a nod to TensorFlow, the open source library for machine learning created by Californians.
A radical change in image capture and processing
To test the power of its chip for the new Pixel 6 and Pixel 6 Pro, Google presented some demos related to image processing. Although there would be very interesting options in the photography section, the juiciest thing would have been seen in the video capture.
The media that accessed the material mention a comparison made when recording the same scene (the sunset on a beach) in HDR, in a 4K setting at 30 frames per second. The mobiles used were the Pixel 6, the Pixel 5 and the iPhone 12 Pro Max , and the first one would have presented the best result.
That the Pixel 6 captures clips with better quality than any previous Pixel is something to note, although it is not surprising; the video section was always the weakest for Google smartphones. But that it is capable of recording equal to or better than a top-of-the-line iPhone , undoubtedly puts the phone on a whole other level. It is necessary to see if the test is reliably transferred to real life in the user experience. But, logically, the feeling reigns that the potential exists.
Osterloh explained that the improvement in the quality of the videos is related to the fact that Tensor is capable of applying the HDRNet process (the same that Google already uses in the photographs) to each frame of the filming .
On the other hand, he also mentioned that they redesigned the image processor and rebuilt the memory architecture to facilitate access to RAM. He even assured that in some sectors they have “directly encoded their image processing algorithms in the hardware”.
Not everything is related to photos and videos
Google made other demonstrations of the power of Tensor on the Pixel 6 that were not related to the photos and videos section. Converting speech to text is another task that promises to greatly benefit from the new SoC developed in Mountain View. For example, the chip is capable of processing the transcription locally , without the need to connect to the internet.
In addition, the hardware can translate in real time from one language to another while playing a video. But that is not all. Voice dictation is now virtually instantaneous, and even allows you to edit words with the keyboard while continuing to speak.
The ability to scale the Tensor SoC beyond Pixels
In the coming months, Google will put a lot of emphasis on the benefits of the Tensor chip and its deep integration with AI and machine learning. Perhaps the biggest challenge is making these features appealing to the general public, and this requires a clever marketing campaign.
However, it is also true to think that the future of the new SoC designed in Mountain View is not necessarily tied exclusively to the family of Pixel smartphones.
Why not think of Tensor as the hardware that will power Chromebooks tomorrow, or a smartwatch with the new WearOS? If Google really commits itself in this way, the potential of its own chips would have no ceiling .
