Tensor G2 vs Google Tensor

We compared two 8-core processors: Google Tensor G2 (with Mali-G710 MP7 graphics) and Google Tensor (Mali-G78 MP20). Here you will find the pros and cons of each chip, technical specs, and comprehensive tests in benchmarks, like AnTuTu and Geekbench.


General comparison of performance, power consumption, and other indicators
CPU Performance
Single and multi-core processor tests
Gaming Performance
GPU performance in games and OpenCL/Vulcan
Battery life
Efficiency of battery consumption
NanoReview Score
Overall chip score

Key Differences

Main differences and advantages of each chip
Pros of Google Tensor G2
  • Announced later
We may receive a commission from Amazon for purchases made using the links below.


Performance tests in popular benchmarks

AnTuTu 9

The AnTuTu Benchmark measures CPU, GPU, RAM, and I/O performance in different scenarios
Tensor G2 +1%
CPU 203616 196221
GPU 295372 297786
Memory 108654 107096
UX 142235 143527
Total score 754442 745917

GeekBench 5

The GeekBench test shows raw single-threaded and multithreaded CPU performance
Single-Core Score
Tensor G2 +1%
Multi-Core Score
Tensor G2 +13%
Image compression 168.3 Mpixels/s -
Face detection 24.5 images/s -
Speech recognition 69.4 words/s -
Machine learning 71.2 images/s -
Camera shooting 31.2 images/s -
HTML 5 2.97 Mnodes/s -
SQLite 1010 Krows/s -


A cross-platform benchmark that assesses graphics performance in Vulkan (Metal)
3DMark Wild Life Performance
Tensor G2 +3%
Stability 71% 55%
Graphics test 38 FPS 37 FPS
Score 6411 6222


Full list of technical specifications of Tensor G2 and Google Tensor


Architecture 2x 2.85 GHz – Cortex-X1
2x 2.35 GHz – Cortex-A78
4x 1.8 GHz – Cortex-A55
2x 2.8 GHz – Cortex-X1
2x 2.25 GHz – Cortex A76
4x 1.8 GHz – Cortex A55
Cores 8 8
Frequency 2850 MHz 2800 MHz
Instruction set - ARMv8.2-A
L2 cache - 3 MB
L3 cache 4 MB 4 MB
Process 5 nanometers 5 nanometers
TDP - 10 W


GPU name Mali-G710 MP7 Mali-G78 MP20
Architecture Valhall Valhall
GPU frequency - 848 MHz
Execution units 7 20
Shading units - 320
FLOPS - 2171 Gigaflops
Vulkan version 1.1 1.1
OpenCL version 2.0 2.0
DirectX version 12 12


Memory type LPDDR5 LPDDR5
Memory frequency 3200 MHz 3200 MHz
Bus 4x 16 Bit 4x 16 Bit
Max bandwidth - 51.2 Gbit/s
Max size 12 GB 12 GB

Multimedia (ISP)

Neural processor (NPU) Next-gen Tensor Processing Unit Tensor Processing Unit
Storage type UFS 3.1 UFS 3.1
Max display resolution 3840 x 2160 3840 x 2160
Video capture 4K at 60FPS 4K at 60FPS
Video playback 4K at 60FPS 4K at 60FPS
Video codecs H.264, H.265, AV1, VP9 H.264, H.265, AV1, VP9
Audio codecs AAC, AIFF, CAF, MP3, MP4, WAV AAC, AIFF, CAF, MP3, MP4, WAV


4G support LTE Cat. 24 LTE Cat. 24
5G support Yes Yes
Download speed - Up to 1200 Mbps
Upload speed - Up to 200 Mbps
Wi-Fi 6 6
Bluetooth 5.2 5.2
Navigation GPS, GLONASS, Beidou, Galileo, QZSS GPS, GLONASS, Beidou, Galileo, QZSS


Announced October 2022 October 2021
Class Flagship Flagship
Model number - S5E9845
Official page - Google Tensor official site

Cast your vote

So, which SoC would you choose?
343 (62.5%)
206 (37.5%)
Total votes: 549


Please give your opinion on the comparison of Google Tensor and Tensor G2, or ask any questions
Ordinary man 09 October 2022 15:46
What!? A76 to A78 with only such differences??
+35 Reply
Paul 23 December 2022 17:50
It's a shame they used a78 cores at all. Two X1 cores will do any job well enough, considering the spare transistor budget could be used for more cache for those cores. But no. This SoC is on the midrange MTK SoC level. Definitely not a flagship.
0 Reply
nunyabizz 08 October 2022 01:10
what a disappointment, using tech from 2018. going the apple route of not focusing on specs, selling a bogus experience
+16 Reply
челты 10 October 2022 15:52
Мощностей даже первого тензора хватает за глаза.
+2 Reply
Вася 06 December 2022 07:20
челты, Ну ну, хватает ему, улучшений ноль.
0 Reply
Paul 14 December 2022 02:31
челты, Ага.. "хватает" это когда 20 ваттная зарядка и паршивая энергоэффективность, которая не улучшилась в сравнении с прошлым поколением. То, что производительности достаточно, не значит что не нужно заботиться о других улучшениях. Использование старой архитектуры ядер явно не пошло на пользу "флагманскому" процессору.
0 Reply