Google shed some light on the Tensor chip during a previous announcement, stating that it will be focused on machine learning, and improve computational photography, but never bothered to provide a silver of information surrounding its performance.
Thankfully, the latest leak mentions that the custom SoC will deliver up to 80% more performance, and while that sounds vague, we are here to clear up the confusion just a little bit.
Google has not provided any direct comparison with the Tensor, but it could be the Pixel 5’s Snapdragon 765G. Let us go through Google’s claim of 80% fast performance.
The 80% performance gain metric is a difference that Google claims was based on internally conducted benchmark testing.
This can mean that the surrounding environment was favorable for the Pixel 6 and the Pixel 6 Pro to perform better while rocking that Tensor chip.
For instance, both flagships could have been placed in a much cooler area, prevented the SoC from reaching its thermal limits, and as a result, performed better.
Google also has not mentioned which chipset it is comparing against that allowed the Tensor to achieve that 80% performance gain. Was it the Pixel 5’s Snapdragon 765G?
We hope that it is because even though the Snapdragon 765G was Qualcomm’s mid-range offering from last year, it could perform in a satisfactory manner, and it sweetened the deal with its embedded 5G modem.
But then again, high-end specs look nice on paper, but you cannot guarantee that they will fare as well when made in practicality. A chip may underperform in synthetic benchmarks but may exceed expectations by having a fast app response time, reducing loading times, and delivering an overall fluid experience.
That difference in app loading times could highlight the Tensor’s performance, and the real magic chip could be due to Google’s software optimization.
After all, an in-house chip should provide more advantages than an ‘off the shelf’ part from Qualcomm.
Unfortunately, we are getting way ahead of ourselves, and we can only consider Google’s claims as legit when the first official Tensor performance results are here.