Google Tensor gets trounced by the A15 Bionic in new Machine Learning benchmark test
We may earn a commission if you make a purchase from the links on this page.
When Google unveiled the Pixel 6 and Pixel 6 Pro last month, it discussed the homemade Google Tensor chip that drives both phones. Google, unlike Apple, failed to mention how many transistors are used by the Tensor (the A15 Bionic sports 15 billion transistors) and instead decided to focus on the component's capabilities with Machine Learning. Unlike the Snapdragon chips used on past Pixel models, Google designed the SoC to work specifically with the computational photography and other features that the Pixel 6 line includes.
You may also want to read:
Apple's A15 Bionic chipset trounced the Google Tensor in an ML-focused benchmark test
Magic Eraser, Motion Mode, Face Unblur are examples of computational photography driven by the Tensor. And the chip also helps the Pixel offer ambient features such as the Pixel 6 series' always-on displays and the Now Playing feature that tells you the name of a song (and the artist) that is playing in the background. The Tensor chip also allows the new Pixels to deliver the most accurate Automatic Speech Recognition (ASR) ever created by Google without consuming too much battery life.
This chart is misleading since the Pixel 5 was powered by the Snapdragon 765G, not a flagship chip
So you would think that when it comes to Machine Learning, the Google Tensor would outperform the A15 Bionic in a benchmark test, but that is not what took place. Instead, on the Geekbench ML test, the Tensor was trounced by Apple's A15 Bionic. The developer of the benchmark test, Primate Labs, says that it is a cross-platform test that is designed to "help you understand whether your device is ready to run the latest machine learning applications."
Grab the Google Pixel 6 and the Pixel 6 Pro
In the Geekbench ML TensorFlow Lite CPU test, the Apple A15 scored 939 compared to just 307 for the Google Tensor. The GPU test was not as lopsided, but Apple's latest chipset still won the battle 2727 (Core ML) while the Tensor scores just 1720 (NNAPI). Overall, the A15 Bionic scored 5,934 on the Geekbenck ML test compared to 3,455 for the Tensor. So even looking at how both chipsets handle Machine Learning tasks, which is supposed to be the strength of the Tensor chip, the A15 Bionic is 71% faster.
So does this mean that you should choose the iPhone 13 Pro Max over the Pixel 6 Pro? No, no it does not. This is a decision that goes beyond the specs. If features like native real-time grammar correction or Live Translate appeal to you more than long battery life or Macro Mode, does it really bother you if the Google Tensor falls short on the Geekbench ML benchmark test?
Do you find that iOS can get boring?
Honestly, iOS can get a little, shall we say, boring. With the quarterly Pixel feature drop, you never know what useful new software Google will be disseminating to your Pixel. And while early indications are that the iPhone 13 Pro Max beats the Pixel 6 Pro in the battery life department, as long as you can get one full day from the Pixel, does it really matter?
Besides, we won't know what to really expect from the Pixel 6 Pro battery for more than a week since it takes a total of two weeks for the adaptive battery feature (which uses Machine Learning) to understand how you use the phone and the battery. And if you need to stretch out the battery on the Pixel 6 and Pixel 6 Pro, disabling 5G will reportedly add more battery life to the handsets.
Remember, most 5G coverage in the U.S. today consists of low-band signals which are not much faster than 4G LTE. The fastest 5G signals are from high-band spectrum which is statistically very hard to find in most areas of the country. Turning off 5G on your Pixel 6 series handset might be worth it if you want the extra battery life.
FCC OKs Cingular\'s purchase of AT&T Wireless
Things that are NOT allowed: