Google Tensor G3 features a Cortex-X3 core, specs leaked

The CPU has a 4 + 4 + 1 core configuration

The Google Pixel 6 was the first smartphone series by the company, which had a chip made by Google. Before that, Google used Qualcomm’s Snapdragon SOCs for its Pixel devices. The in-house chipset was named “Tensor,” this SOC was also used in the Google Pixel 6a, the budget Pixel of 2022.

The first Tensor chip had a 4 + 2 + 2 core configuration, including two Cortex-X1 cores clocked at up to 2.8GHz, two Cortex-A76 cores clocked at up to 2.25GHz, and four Cortex-A55 cores clocked at up to 1.8GHz. The Google Tensor G2 was launched with the same 4 + 2 + 2 configuration and even had the same two Cortex-X1 cores, this time clocked at 2.85GHz. There were two Cortex-A78 cores clocked at 2.3GHz and four Cortex-A55 cores clocked at 1.8GHz.

Do note that we are talking about a chip launched in 2022 when we had other chipsets from Qualcomm & MediaTek featuring the more powerful Cortex-X2, Cortex-A710, and Cortex-A510 cores.

Google Tensor G3

But this year, Google is bringing some significant upgrades to the Tensor G3 chipset, as per a report by Android Authority.

The most significant upgrade is the core configuration, as Google Tensor G3 features a 4 + 4 + 1 configuration for the CPU. There is a prime Cortex-X3 core clocked at up to 3.0GHz, four Cortex-A715 cores at up to 2.45GHz, and four Cortex-A510 cores at up to 2.15GHz. This should significantly improve the CPU performance over the Tensor G2.

As for the GPU performance, the Google Tensor G3 features a 10-core ARM Mali-G715, aka the ARM Immortalis. The maximum GPU frequency is 890MHz. The last year’s Tensor G2 had a 7-core Mali-G710 GPU with a maximum frequency of 848MHz. The ARM Immortalis GPU is the first GPU offering hardware acceleration for ray tracing and is the GPU used in the MediaTek Dimesity 9200.

The main USP of the Google Tensor chipsets is not the CPU performance or power efficiency but rather the AI performance. Google Tensor chipsets have a built-in TPU or Tensor Processing Unit, which handles the AI tasks. The first Tensor chip had a TPU codenamed “Abrolhos” clocked at 1.0GHz. The Tensor G2 got an upgraded TPU, clocked at 1.0GHz, and offered up to 60% better performance than the previous generation. With the Tensor G3, Google upgraded the TPU, codenamed “Rio,” clocked at 1.1GHz.

According to the source, the Google Tensor G3 chipset supports faster UFS 4.0 storage, has an upgraded 1065MHz GXP codenamed “callisto”, and is the first smartphone chip to support AV1 encode.

All these details suggest a massive performance, efficiency, and AI upgrade in the Google Pixel 8 series. It is also worth mentioning that Qualcomm is yet to launch the Snapdragon 8+ Gen 2 chip & MediaTek is already leading the charts with its Dimensity 9200+ SOC. Let’s see how the Google Tensor G3 competes against the best that Qualcomm & MediaTek have to offer because, as we know, in the last two years, there hasn’t been competition at all since the Snapdragon & Dimensity chipsets have been much better than the Tensor chips in CPU & GPU performance.