There might have been a collective sigh of relief when Google announced its upcoming self-designed Tensor processor and Pixel 6 phones. It seemed that, after a hiatus, Google is ready to once again turn the Pixel into a premium smartphone brand, both in looks and in specs. The Tensor chip, however, has so far been wrapped in mystery and machine learning buzzwords, language that will likely confound most consumers asking the simple question of whether it’s powerful enough or not. The Pixel 6 Pro has finally surfaced in a benchmark to provide that answer, but it’s still too premature to draw any conclusions from it.
Google has been unsurprisingly secretive about the technical aspects of the Tensor System-on-Chip (SoC), aside from its high-level benefits when it comes to AI and ML processing. There have been a few rumors here and there, but nothing as substantial as this Geekbench entry. We’re finally getting a glimpse at what the Google Tensor might be like, and, suffice it to say, it’s not your usual mobile AP (application processor).
For one, the configuration of cores is unconventional or at least rare. There are two cores running at 2.80GHz, two at 2.25GHz, and four at 1.80 GHz. Given it’s not likely that the Tensor will have two powerful Cortex-X1 cores (which is already last-gen technology by now), we might be looking at four Cortex-A78 cores grouped into two and four Cortex-A55 cores. The configuration does at least seem to be on par with something like a Snapdragon 888, which makes the benchmark scores even stranger.
The Pixel 6 Pro scored a measly 414 in Geekbench’s single-core test and 2074 in the multi-core suite. That’s considerably lower than the average score for the Snapdragon 765G running in the Pixel 5a (5G), much less a flagship Snapdragon 888, but there is an important catch. At least according to speculation, this Pixel 6 Pro might be a pre-production unit running on pre-release software, suggesting that the final product will hopefully be up to snuff.
The benchmark does give the Pixel 6 Pro 12GB of RAM, putting it on the same footing as most premium smartphones this year. It will definitely be interesting to see how the Tensor chip performs in the real world and whether Google’s machine learning expertise is really enough to compete with Apple’s long history of making its own silicon.