Qualcomm Snapdragon 665 vs HiSilicon Kirin 960 vs HiSilicon Kirin 970

Qualcomm Snapdragon 665

The Qualcomm Snapdragon 665 (SM6125) is a fast mid-range ARM-based SoC largely found on Android tablets and smartphones. It features 8 Kryo 260 cores (custom design, 64-Bit capable) that are divided in two clusters. A fast cluster of four cores with up to 2 GHz (Kryo 260 Gold - Cortex-A73 derivate) and a power saving efficiency cluster with up to 1.8 GHz (Kryo 260 Silver - Cortex A53 derivate). Both clusters can also be used together.

In addition to the 8 CPU cores, the SoC integrates a mid range Adreno 610 GPU with a LPDDR4 memory controller (dual-channel)  and supports Wi-Fi (802.11ac + MIMO, max. 867 Mbps), Bluetooth 5, and LTE (X12 LTE modem with up to 600 Mbps download and 150 Mbps upload).

Power Consumption

Thanks to the new 11 nm LPP process, the energy efficiency should be quite good.

HiSilicon Kirin 960

The HiSilicon Kirin 960 is an ARM based high-end octa-core SoC for smartphones and tablet, which was introduced with the Huawei Mate 9 in November 2016. Besides 8 CPU cores (4x Cortex-A73, 4x Cortex-A53), the SoC is also equipped with a modern Mali-G71 MP8 graphics adapter, a dual-channel LPDDR4 memory controller as well as an LTE Cat. 12/13 modem. It is one of the fastest ARM SoCs in the end of 2016. The Kirin 960 is manufactured in a modern 16-nm process at TSMC.

HiSilicon Kirin 970

The HiSilicon Kirin 970 is an ARM based high-end octa-core SoC for smartphones and tablet, which was introduced with the Huawei Mate 10 in late 2017. Besides 8 CPU cores (4x Cortex-A73, 4x Cortex-A53), the SoC is also equipped with a modern Mali-G72 MP12 graphics adapter, a dual-channel LPDDR4 memory controller as well as an LTE (4x4 MIMO, 5-band-carrier-aggregation, QAM-256, up to 1.2 Gbit/s) modem.

The performance cluster is clocked at up to 2.4 GHz and the four A53 power saving cores with up to 1.8 GHz. The performance should improve by about 20 percent according to Huawei and the power consumption reduced by 20 percent.

On the chip HiSilicon also included a dedicated Neural Processing Unit (NPU) for running machine learning tasks very efficiently compared to the CPU cores.