Tech —

The Pixel 2’s custom camera SoC uses Intel technology

Google had a little help building its first-ever consumer SoC.

The Pixel 2’s custom camera SoC uses Intel technology
Ron Amadeo/Intel

Google's Pixel 2 smartphone doesn't just have one of the best smartphone cameras ever; it also has custom silicon dedicated to the camera that isn't even active yet. Besides the Snapdragon 835, the Pixel 2 has a whole other SoC for image processing called the "Pixel Visual Core." The chip represents Google's first-ever consumer SoC, but Google didn't build the chip on its own. CNBC found out the chip was a collaboration between Intel and Google.

CNBC made the connection after seeing that the serial number of the chip starts with "SR3," which is also used on some Intel chips. The outlet ran its scoop by Google, which confirmed Intel was involved.

Knowing that Intel helped with the development of the chip was enough information to start digging with, since anything touched by Intel is probably related to the camera chip, right? This led me to the codeword "Easel," which, sure enough, seems to be Google's codename for the Pixel Visual Core. You can poke around platform/hardware/google/easel/ in the Android source, where you'll find the few bits of related code that are currently public. Opening up the device-tree blob binary present on the Pixel 2 also prominently shows the word "Monette Hill," which sounds like some kind of Intel codename.

A smartphone design win for Intel is a rare occurrence, since the industry's reliance on ARM processors means Intel is usually absent from the world's most popular computing form factor. The company has made inroads on smartphone modems, which show up in certain iPhones. For Android OEMs, a separate Intel modem is a tough sell when Qualcomm can offer modems integrated with its SoCs.

Google's Pixel Visual Core isn't active yet, but the company says it will be turned on with the launch of Android 8.1. The 8-core Image Processing Unit (IPU) will supposedly allow Google's HDR+ image processing to run "5x faster and at less than 1/10th the energy" of the current CPU-driven implementation. It will also be a programmable platform for other Google imaging and machine-learning functions.

Channel Ars Technica