Google debuted the Tensor chip in August along with the Pixel 6 and Pixel 6 Pro, but Tuesday was the first time the company has talked in-depth about the custom-built SoC. With Tensor, things like speech, language, imaging, and video on Pixel phones are now heterogeneous, Google said. This means they use multiple resources across the entire chip. The chip can run more advanced, state-of-the-art machine learning at lower power consumption rates compared to previous Pixel phones. It also uses the most accurate Automatic Speech Recognition (ASR) Google’s ever released, which means ASR can now work with applications that run for a long time, such as Recorder and Live Caption, without wiping out your battery. Tensor also promises to make Google’s Live Translate feature better. Pixel 6 owners can use chat apps like Messages and WhatsApp to translate foreign languages directly in those apps, instead of cutting and pasting text into Google Translate. Live Translate now also works on videos. Pixel phones are known for having pretty impressive cameras, too, and Tensor should help make your snaps even prettier. Its architecture means the Pixel 6 and Pixel 6 Pro can capture images more quickly, Google said, and it powers new features like Motion Mode and HDRNet (image enhancement). Now that HDRNet is embedded directly onto the chip, it works in all video modes, which is a first for Google. Face recognition is more accurate and consumes less power, as well. The new Tensor chip also brings improvements to hardware security. Tensor comes with a CPU-based subsystem that works with the Titan M2 security chip to protect data. “Independent security lab testing showed that Titan M2 can withstand attacks like electromagnetic analysis, voltage glitching, and even laser fault injection. Yes, we literally shot lasers at our chip!” Monika Gupta, the senior director of Google Silicon, wrote in a blog post. The Pixel 6 and Pixel 6 Pro launch on October 28 and will cost $599 and $899, respectively.
