Google Tensor G2

Google Tensor G2

Google Tensor G2: Deep Integration of AI and Performance Balance

April 2025

Since the launch of the first Tensor in 2021, Google has been focusing on the unique capabilities of artificial intelligence in its processors. Tensor G2, introduced in 2023, represents the evolution of this strategy, offering improved performance, energy efficiency, and deep integration with the Android ecosystem. In this article, we will explore how this chip compares to its competitors, who it is suitable for, and which tasks it excels at.


1. Architecture and Process Technology: What's Inside Tensor G2?

CPU: 8-Core Hybrid Structure

The Tensor G2 employs a hybrid architecture with 8 cores divided into three clusters:

- 2 high-performance Cortex-X1 cores with a clock speed of up to 2.85 GHz for resource-intensive tasks.

- 2 Cortex-A78 cores (mid-range) running at 2.35 GHz.

- 4 energy-efficient Cortex-A55 cores (1.8 GHz) for background processes.

This approach (big.LITTLE++) allows for flexible task distribution, saving battery life. However, compared to the Snapdragon 8 Gen 2 (1+4+3) or Apple A16 (2+4), the Tensor G2 appears conservative—Google prioritizes not maximum frequencies but optimization through AI.

GPU: Mali-G710 MP7

The Mali-G710 MP7 GPU supports ray tracing in mobile games and improved video processing. In 3DMark Wild Life Extreme tests, it is 20% faster than the Mali-G78 in Tensor G1, but lags behind the Adreno 740 in Snapdragon 8 Gen 2 by 15-25% in FPS.

Process Technology: 5 nm (Samsung)

The chip is fabricated using Samsung's 5nm process technology, providing a balance between performance and a TDP of 7 watts. However, modern competitors (like Snapdragon 8 Gen 3) have already transitioned to 4nm, giving them an advantage in energy efficiency under load.


2. Real-World Performance: Where Does Tensor G2 Shine?

Gaming

The average performance in AAA titles (e.g., Genshin Impact) is 50-55 FPS on medium settings. In comparison, Snapdragon 8 Gen 2 consistently delivers 60 FPS on high settings. However, Tensor G2 experiences less overheating due to aggressive throttling—after 20 minutes of gaming, the GPU frequency drops by 10-15%, but the temperature stays within 42-44°C.

Multimedia and AI

- Video: The chip decodes 8K at 30fps and encodes 4K at 60fps with HDR+.

- Photo: Processing images in the Pixel 8 Pro (powered by Tensor G2) takes 30% less time than on Tensor G1, thanks to the TPU v3 AI block. For instance, the Magic Eraser feature removes objects from photos in 1-2 seconds.

- Energy Consumption: With Always-On Display and moderate use (social media, streaming video), the smartphone lasts 1.5-2 days. Under load (gaming, video recording), the 4500 mAh battery drains in 4-5 hours.


3. Built-in Modules: Connectivity and More

- Modem: The integrated Exynos 5300 supports 5G mmWave and Sub-6, LTE Cat.24 (up to 3 Gbps). However, in weak signal areas, the Snapdragon X70 (in Snapdragon 8 Gen 2) shows 15-20% better stability.

- Wi-Fi 6E and Bluetooth 5.3: Supports tri-band Wi-Fi and LE Audio for quality sound in TWS earbuds.

- Navigation: GPS, GLONASS, Galileo, QZSS. Notably, there is improved integration with the Emergency SOS via Satellite service in the Pixel 8 Pro—sending SOS messages even without cellular connectivity.


4. Comparison with Competitors: Who Are the Leaders?

Google Tensor G2 vs Snapdragon 8 Gen 2 vs Apple A16 Bionic

- Geekbench 6: Tensor G2 (Single-Core: 1188, Multi-Core: 3468) falls behind the Snapdragon 8 Gen 2 (1305/4200) and A16 (1880/5450) due to less powerful cores.

- AnTuTu 10: 824,434 points compared to 1,200,000 for Snapdragon 8 Gen 3 (2024) and 950,000 for A16.

- AI Benchmarks: In MLPerf tests, Tensor G2 outperforms competitors in tasks related to NLP (natural language processing) due to optimization for Google Assistant and Bard.

Conclusion: Tensor G2 is not a top-tier chip for gamers but is ideal for scenarios involving active use of AI and streaming.


5. Use Cases: Where Does the Processor Shine?

- Everyday Tasks: Smooth operation in Chrome with 20+ tabs open, switching between messengers, streaming in 4K.

- Photo and Video: Modes like Night Sight, Super Res Zoom, and Cinematic Blur work in real time. For instance, recording 4K at 60fps while simultaneously applying HDR processing and stabilization.

- Gaming: Suitable for casual games (COD Mobile, Asphalt 9) on high settings. For heavier projects, graphics will need to be reduced.

- AI Applications: Local execution of tasks in apps like Adobe Lightroom Mobile (AI Denoise) or Google Recorder (offline transcription).


6. Pros and Cons

Strengths:

- Best-in-class integration with Google services and AI.

- Stable operation without overheating during extended sessions.

- Support for current communication standards (Wi-Fi 6E, 5G mmWave).

Weaknesses:

- Lagging in GPU performance compared to Snapdragon and Apple.

- Lack of hardware acceleration support for AV1 (important for YouTube 8K).

- High device prices (Pixel 8 Pro starts at $999).


7. Practical Tips: How to Choose a Smartphone with Tensor G2?

- Cooling: Devices with a vapor chamber (e.g., Pixel 8 Pro) are less prone to throttling.

- RAM: At least 8 GB of RAM for comfortable use with AI features.

- Price: In 2025, smartphones with Tensor G2 (like Pixel 8a) are available from $599. Flagships range from $799 to $999.

Typical Devices:

- Google Pixel 8/8 Pro — for fans of pure Android and photography.

- Xiaomi T-Series (Chinese market) — budget models focused on AI.


8. Final Conclusion: Who Is Tensor G2 For?

This processor is designed for those who value:

- AI Capabilities — quick photo processing, smart assistants, offline voice transcription.

- Stability — smooth system performance even after 2-3 years.

- Energy Efficiency — long battery life.

It is not meant for hardcore gamers but is ideal for content creators, travelers, and users looking for a "smart" device with an emphasis on camera and Google ecosystem integration. The primary benefit is the harmony between software and hardware that is hard to find in other Android smartphones.

Basic

Label Name
Google
Platform
SmartPhone Flagship
Launch Date
October 2022
Manufacturing
Samsung
Model Name
Tensor G2
Architecture
2x 2.85 GHz – Cortex-X1 2x 2.35 GHz – Cortex-A78 4x 1.8 GHz – Cortex-A55
Cores
8
Technology
5 nm
Frequency
2850 MHz

GPU Specifications

GPU name
Mali-G710 MP7
GPU frequency
850 MHz
FLOPS
1.1424 TFLOPS
Shading units
96
Execution units
7
Max display resolution
3840 x 2160

Connectivity

4G support
LTE Cat. 24
5G support
Yes
Bluetooth
5.2
Wi-Fi
6
Navigation
GPS, GLONASS, Beidou, Galileo, QZSS, NAVIC

Memory Specifications

Memory type
LPDDR5
Memory frequency
3200 MHz
Memory Bus
4x 16 Bit
Max Bandwidth
51.2 Gbit/s

Miscellaneous

Neural processor (NPU)
Next-gen Tensor Processing Unit
Audio codecs
AAC, AIFF, CAF, MP3, MP4, WAV
Max camera resolution
1x 200MP, 2x 32MP
Storage type
UFS 3.1
Video capture
4K at 60FPS
Video codecs
H.264, H.265, AV1, VP9
Video playback
4K at 60FPS
TDP
7 W
Instruction set
ARMv8.4-A

Benchmarks

Geekbench 6
Single Core Score
1188
Geekbench 6
Multi Core Score
3468
Geekbench 5
Single Core Score
584
Geekbench 5
Multi Core Score
1783
FP32 (float)
Score
1119
AnTuTu 10
Score
824434

Compared to Other SoC

Geekbench 6 Single Core
3842 +223.4%
1196 +0.7%
1188
471 -60.4%
288 -75.8%
Geekbench 6 Multi Core
14383 +314.7%
3468
2331 -32.8%
1509 -56.5%
866 -75%
Geekbench 5 Single Core
897 +53.6%
623 +6.7%
584
413 -29.3%
309 -47.1%
Geekbench 5 Multi Core
2413 +35.3%
2142 +20.1%
1783
1229 -31.1%
A9
962 -46%
FP32 (float)
6110 +446%
1788 +59.8%
1119
588 -47.5%
392 -65%
AnTuTu 10
3518353 +326.8%
1135053 +37.7%
824434
505728 -38.7%
406460 -50.7%