Google debuts custom-made Tensor chips: All about the powerhouse inside the Pixel 6

Google, in its Pixel Fall event on October 19, has introduced a brand new 'Tensor' chip that will power the much-awaited Pixel 6 and Pixel 6 Pro. Designed especially for the Pixel devices, Tensor is a custom-made chip that will boost the smartphone's performance by 80% as compared to Pixel 5's Qualcomm Snapdragon 765G processor. Google promises that the Tensor chip will also offer incredible machine learning that will help the users to translate messages and videos without the internet. 

Reflecting on the processor, Google's Senior Vice President Rick Osterloh said, “Google Tensor is the biggest mobile hardware innovation in the history of the company. It’s the culmination of years of investment in AI and Google’s deep experience in silicon.”

Specialties of Google Tensor

The Tensor chip is a result of the collaboration of Google engineers across hardware, software and machine learning. The chip is divided into seven sections each of which performs a different function. The top section is the integrated ML engine called the 'TPU' that is custom-made by Google research and for Google research. The second section on the top called the 'Image Signal Processor' is responsible for the power efficiency of Pixel devices. The Pixel device's CPU and GPU are also designed to complement the machine learning that will offer advanced computational operations.

The chip has a CPU with a 2+2+4 configuration, which means it has two high-performance cores, two mid cores and four high-efficiency cores. Besides the Pixel, devices will also sport a 20-core GPU for the premium gaming experience. 

Another section belongs to the Context Hub that brings machine learning to an ultra-low-power domain, which ensures seamless functioning without draining the battery. 

With the Tensor chip in Pixel devices, Google has also ensured unhindered heterogeneous computing i.e. when certain applications on a smartphone use multiple parts of the chip. Now with Tenors in place, the device will use different subsystems of the chip rather than just optimising individual elements for peak performance. 

Moreover, the processor also consists of two other units called the system cache and security that add to its enhanced performance.

Image: Twitter/@Benwood