We benchmark the Google Pixel 6 Tensor graphics, and the scores impress
We had to test our own Pixel 6 and Pixel 6 Pro units to confirm what's been appearing in benchmark databases lately, but the new Google Tensor chipset GPU indeed seems better than what both the Snapdragon 888 and Exynos 2100 processors in phones like the Galaxy S21 Ultra offer.
Google's homebrew Tensor chipset uses a tailored 20-core graphics processing subsystem that it says is 370% more powerful than what's in the Pixel 5 specs, in addition to a dedicated image processing core that is used exclusively for the photography algorithms.
Google Tensor vs Snapdragon 888 GPU chip benchmark
We ran the graphics-intensive Wild Life and Wild Life Extreme tests, as well as the CPU-centered Geekbench, and the really impressive scores came from the Google Pixel 6 Tensor GPU.
That's understandable, considering it has to push about two million pixels less than the Pixel 6 Pro graphics subsystem, yet even the Pro's 6000+ result is better than what both the Snapdragon 888 and Exynos 2100 GPUs are able to pull off on average in the same test.
When it comes to sustained performance in the Wild Life Extreme test, the Pixel 6 and 6 Pro throttle with the best of them when they get hot and bothered under pressure, but the scores are still rather respectable. Moreover, they come after running the single Wild Life test, so throttling has probably been worse than if Tensor started cool as a cucumber from the get-go.
The overall Tensor chip benchmark score in Geekbench is also very good in the single core tally, thanks to the powerful X1 core, of which it has two compared to one for Snapdragon 888 or Exynos 2100. In multi-core performance the results give way a bit, though, as Google decided to go with the Cortex-A76 architecture for Tensor's two midrange cores instead of the A78 that's in its rivals.
Google Pixel 6 Pro Tensor vs Galaxy S21 Ultra Snapdragon 888 benchmark scores
All of this comes to show that Google has optimized Tensor for peak performance, 3D graphics and image processing, as well as AI calculations, rather than raw chipset benchmark numbers, and we can't wait to test the Google Pixel 6 and 6 Pro battery life to determine if that strategy is as power-efficient as it is task-driven.
Things that are NOT allowed: