Google Tensor 2 will be the Pixel 7 SoC, muscle or brain?

Just yesterday we confirmed the presentation date of the Pixel 7, a high-end smartphone that will be the star of an event that Google is preparing for October 6. As usual, we will be attentive to tell you all the keys of said terminal, although we can already advance some important details about its SoC, the Google Tensor 2.

I am sure that many of our readers will remember that with the Pixel 6, the Mountain View company decided to take a major turn at the hardware level. Google ditched Qualcomm’s Snapdragon series, opting for mount a “own” chip known as a Tensor, in quotes because it was actually a Samsung Exynos with a slight layer of customization.

We already told you that the Google Tensor was a high-performance chip that is based on a Samsung Exynos design, and that it has a 8 core CPU divided into two X1 cores at 2.8 GHz, two Cortex-A76 cores at 2.25 GHz and four low-power cores at 1.8 GHz. Its GPU is a Mali-G78MP20, it has a Titan M2 security processor and a TPU for AI and deep learning. It is manufactured in 5nm.

I wanted to share with you the keys to the original Google Tensor because, in general, everything seems to indicate that the second generation model is going to follow in its footsteps. This means that the Mountain View giant is going to prioritize brains over brawn, that is, the specialization and value of AI and deep learning over raw performance. This makes a lot of sense, and is perfectly understandable, especially after seeing how key technologies such as computational photography supported by artificial intelligence make a difference.

Does this mean that Google Tensor 2 is going to be mediocre in performance? Not at all, but it does imply that it will not be at the same level than a current top-of-the-range chip, at least in raw power. This does not have to be a problem, since today’s upper-middle-range chips are already “exceeding” in almost any task, and offer excellent performance even with demanding applications.

So what can we expect from Google Tensor 2? Well, a performance improvement at the CPU and GPU level that places it at the level of current high-end SoCs, and we could also see a new TPU and a new ISP (image signal processor for its acronym in English). It should replicate the 5nm fabrication node, and it will most likely give us access to exciting new features supported by AI.