An in-depth analysis reveals all the secrets of Google Tensor

As we all know since before the announcement of Google Pixel 6 and Pixel 6 Pro, the new high-end smartphones of the Mountain View giant are the first to mount the company’s first proprietary processor: Google Tensor.

An in-depth analysis on Google Tensor

In these hours, the colleagues of AnandTech have published an in-depth analysis of Google Tensor putting on paper some features that have been circulating on the net for some time. First, the chip Google shares many similarities with the processor family Exynos of Samsung: same power management architecture and memory, read the analysis.

Google Tensor and Samsung Exynos 2100: Similar But Different
SoC

Google Tensor

Exynos 2100

CPU 2x Cortex-X1
@ 2.80GHz 2x1024KB pL2

2x Cortex-A76
@ 2.25GHz 2x256KB pL2

4x Cortex-A55
@ 1.80GHz 4x128KB pL2

4MB sL3

1x Cortex-X1
@ 2.91GHz 1x512KB pL2

3x Cortex-A78
@ 2.81GHz 3x512KB pL2

4x Cortex-A55
@ 2.20GHz 4x64KB pL2

4MB sL3

GPU Mali G78 MP20 @
848 MHz (shaders)
996 MHz (tiler / L2)
Mali G78 MP14 @
854 MHz
Memory
Controller
4x 16-bit CH

@ 3200MHz LPDDR5  /  51.2GB/s

8MB System Cache

ISP Hybrid Exynos + Google ISP Full Exynos ISP Blocks
+ DSP
NPU Google edgeTPU
1066MHz
Exynos NPU
@ 1352MHz
Media Samsung Multi-Function Codec

8K30 & 4K120 encode &
8K60 decode

H.265/HEVC, H.264, VP9
AV1 Decode

Google “BigOcean”
4K60 AV1 Decode
Modem Exynos Modem 5123
External

(LTE Category 24/18)
(5G NR Sub-6)
(5G NR mmWave)

Exynos Modem 5123
Integrated

(LTE Category 24/18)
(5G NR Sub-6)
(5G NR mmWave)

Mfc. Process Samsung
5nm (5LPE)

However, Google has managed some rather fundamental aspects of the new processor in a different way. First of all, the setup of the cores are very different from the typical ones used by Samsung on the Exynos: the US company has in fact adopted a “2 + 2 + 4” configuration, and it is not clear the reason behind the implementation of the Cortex-A76 cores. now dated and with poor performance management.

Join RealMi Central on Telegram, Facebook & Twitter

The idea is that Google, with the Tensor chip, had a mixed approach: on the one hand it relied on part of Samsung’s Exynos chip architecture, on the other hand, it implemented some customizations that prompted AnandTech to describe the processor as semi-custom. If you want to know more information on Google’s processor, consumption, performance and much more, we leave you the link in the source.

Leave a Comment