When it announced the Pixel 7 series phones last week, Google also took a moment to introduce the second custom processor Tensor G2 in cooperation with Samsung, emphasizing that it can bring faster artificial intelligence operation response, and the Sammobile website further obtains. Details of this processor design.
Among them, Tensor G2 is produced by Samsung’s 4nm LPE process technology. Through the “2+2+4” three-cluster architecture configuration, two sets of Arm Cortex- X1 CPUs with an operating clock of 2.85GHz are used, and two sets of operating clocks are used. It is a Cortex-A78 CPU of 2.35GHz, and is also equipped with a Cortex-A55 CPU with an operating clock of 1.8GHz, and a Mali-G710 GPU designed with 7 sets of graphics cores.
As for the 5G networking capability, Tensor G2 integrates Samsung Exynos 5300 5G networking data chip, TPU tension processor, and Titan M security chip, and distributes computing execution efficiency through Context Hub design.
The Sammobile website further pointed out that Google will continue to cooperate with Samsung to customize the processor, such as using Samsung’s new 3nm GAA process technology to improve processor computing performance and reduce power consumption, but Google does not rule out cooperation with TSMC for advanced processes technology, so that its customized processors can have higher computing performance, and at the same time can compete with the computing performance of mainstream processors in the market.
Prior to this, Google emphasized that the purpose of investing in customized processors is to achieve vertical integration of software and hardware, and maximize computing efficiency, while improving power usage efficiency to avoid mainstream processors with higher computing performance. The actual use process produces higher execution efficiency, but leads to more power loss and waste, and even leads to excess performance.
Therefore, in its custom processor design concept, it will not aim to improve computing performance, but to improve the computing performance of artificial intelligence as much as possible, such as greatly increasing the accuracy of speech recognition, significantly improving recognition speed, and extending battery life time.