Pixel, the predecessor to the Nexus S, has released the Pixel 6 and Pixel 6 Pro models. These are both powered by a non Qualcomm processor, google’s own tensor processors. Google’s big effort on Tensor, which is built on the AI-boosting TPU and promises to improve photographs and videos, search, captioning, text-to-speech, and more. It’s a tough goal for any chip, let alone someone whose biggest selling point is machine learning. So what’s the performance of Google’s Tensor chip in the Pixel 6? Does this dedication to machine learning-powered features come at the expense of overall performance?
It's not about the Benchmark score
Despite its primary eight-core CPU, the Google Tensor’s 5nm architecture is energy efficient. The CPU is made up of two 2.8GHz Cortex-X1 cores for demanding work, two 2.25GHz Cortex-A76 high-performance cores, and four 1.8GHz Cortex-A55 cores for routine tasks.
Tensor appears to lag behind the Snapdragon 888-powered competition in this area, with single- and multicore scores finishing off the pace established by rival Android flagships. The fastest Android phone in this test, the RedMagic 6S Pro gaming phone powered by the Snapdragon 888 Plus, has a single core score that is 10% better than the Pixel 6 Pro, and a multicore score that is 34% higher.
The phone’s 50MP ISP claims minimal shutter latency and Google’s HDRnet, while a 20-core ARM Mali GPU handles computational photography and gaming. Context Hub is also included, which is a coprocessor that handles with contextual reminders and the always-on display. Google has included lots of new security coprocessors in the Pixel 6 and Pixel 6 Pro. Tensor security core is a separate CPU subsystem that handles your most sensitive processes. It interacts with Google’s Titan M2 chip to safeguard passcodes, encrypt data, and protect sensitive data shared online.
The Tensor chip from Google is unlikely to dominate benchmarks, but that’s not the point. Google intended silicon in which everything works together to provide maximum efficiency, performance, and security. While certain smartphone benchmarks give useful and verifiable information, they are often incorrect projections of everyday performance.
Generally, Google’s Tensor SoC within the Pixel 6 and Pixel 6 Pro looks to be a significant first step for Google in collaboration with Samsung. While it remains to be seen where Google goes with the Tensor SoC in the future and how much more tweaking the firm makes, it is evident that the business has a flagship-class SoC that comes close to defeating the competition in certain areas while winning in others. If Google’s objective was to create the fastest processor for machine learning on an Android device, it appears that they have succeeded.
The performance disparity between Qualcomm’s Snapdragon processors raises the question of whether the Tensor project’s purpose was truly performance-oriented or if it was more concerned with cost and roadmap control. If Google doesn’t ship enough Pixels or other devices with the Tensor SoC, I’m not sure how much of a cost-cutting exercise building a semi-custom chip would be.