Home > Smartphone processors comparison > Tensor G6 vs Tensor G5 – what's better?

Tensor G6 vs Tensor G5

Tensor G6
VS
Tensor G5
Tensor G6
Tensor G5

We compared the 7-core Google Tensor G6 (PowerVR CXTP-48-1536) with the older 8-core Tensor G5 (PowerVR DXT-48-1536) SoC. Here you will find the pros and cons of each chip, technical specs, and comprehensive tests in benchmarks, like AnTuTu and Geekbench.

  1. Review
  2. Differences
  3. Benchmarks
  4. AnTuTu v11
  5. GeekBench 6
  6. Specs
  7. Comments (17)

Review

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/Vulkan
Battery life
Efficiency of battery consumption
NanoReview Score
Overall chip score

Key Differences

Main differences and advantages of each chip
Pros of Google Tensor G6
  • Higher GPU frequency (~17%)
  • Performs 17% better in floating-point computations
  • Shows better (up to 25%) AnTuTu 11 score – 1933K vs 1549K
  • 9% higher CPU clock speed (4110 vs 3780 MHz)
  • Better instruction set architecture
Pros of Google Tensor G5
  • Has 1 more cores

Benchmarks

Performance tests in popular benchmarks
SoC:
Tensor G6
vs
Tensor G5

AnTuTu 11

The AnTuTu Benchmark measures CPU, GPU, RAM, and I/O performance in different scenarios
Tensor G6 +25%
1933536
Tensor G5
1549798
CPU 876037 734567
GPU 288736 183228
Memory 304202 222379
UX 464561 409624
Total score 1933536 1549798
Samples [4 tests] [16 tests]
Submit your AnTuTu result

GeekBench 6

The GeekBench test shows raw single-threaded and multithreaded CPU performance
Single-Core Score
Tensor G6 +20%
2755
Tensor G5
2305
Multi-Core Score
Tensor G6 +25%
7563
Tensor G5
6054
Asset compression - 281.8 MB/sec
HTML 5 Browser - 162.5 pages/sec
PDF Renderer - 207.4 Mpixels/sec
Image detection - 149.5 images/sec
HDR - 151.4 Mpixels/sec
Background blur - 24.4 images/sec
Photo processing - 46.5 images/sec
Ray tracing - 6.79 Mpixels/sec
Samples 6 tests 8 tests
Compute Score (GPU)
Tensor G6 +51%
6216
Tensor G5
4116
Background blur - 4.77 images/sec
Face detection - 2.41 images/sec
Horizon detection - 0.17 Gpixels/sec
Edge detection - 0.27 Gpixels/sec
Gaussian blur - 0.2 Gpixels/sec
Feature matching - 0.05 Gpixels/sec
Stereo matching - 12.6 Gpixels/sec
Particle physics - 634.9 FPS
API - Vulkan
Samples 3 tests 7 tests

3DMark

A cross-platform benchmark that assesses graphics performance in Vulkan (Metal)
3DMark Steel Nomad Light
Tensor G6
n/a
Tensor G5
1012
Score - 1012
Stability - 96%
Graphics test - 7 FPS

PCMark 3.0

Benchmark for evaluating overall system performance in everyday workloads
PCMark 3.0
Tensor G6
n/a
Tensor G5
15332
Web score - 17237
Video editing score - 7203
Data manipulation score - 14036
Photo editing score - 24584
Writing score - 20213
Phones tested - 4 models

Recent User Tests

The latest benchmark tests that have been submitted by users
Google Tensor G6
DateBenchmarkResult
πŸ“˜ 2026-08-23 (Roni)Geekbench 6 CPU2888 / 7641
πŸ“˜ 2026-08-14 (Flxen)Geekbench 6 CPU2755 / 7563
Add View all (2)
Google Tensor G5
DateBenchmarkResult
πŸ“˜ 2026-08-22 -> PixeluserAnTuTu 111536042
πŸ“˜ 2026-08-02 -> bangAnTuTu 111719659
πŸ“˜ 2026-08-02 -> bangCPU Throttle74%
πŸ“˜ 2026-07-31 -> bangGeekbench 6 GPU4116
πŸ“˜ 2026-07-31 -> bangCPU Throttle72%
Add View all (22)

Specifications

Full list of technical specifications of Tensor G6 and Tensor G5

CPU

Architecture 1x 4.11 GHz – C1-Ultra
4x 3.38 GHz – C1-Pro
2x 2.65 GHz – C1-Pro		 1x 3.78 GHz – Cortex-X4
5x 3.05 GHz – Cortex-A725
2x 2.25 GHz – Cortex-A520
Cores 7 8
Frequency 4110 MHz 3780 MHz
Instruction set ARMv9.3-A ARMv9.2-A
L3 cache 8 MB 8 MB
Process 3 nanometers 3 nanometers
Manufacturing TSMC TSMC

Graphics

GPU name PowerVR CXTP-48-1536 PowerVR DXT-48-1536
GPU frequency 1286 MHz 1100 MHz
Pipelines 6 6
Shading units 128 128
Total shaders 768 768
FLOPS 1975.3 Gigaflops 1689.6 Gigaflops
Vulkan version 1.4 1.4
OpenCL version 3.0 3.0

AI Accelerator

Neural processor (NPU) Yes Yes

Memory

Memory type LPDDR5X LPDDR5X
Memory frequency 4266 MHz 4266 MHz
Bus 4x 16 Bit 4x 16 Bit
Max bandwidth 68.2 Gb/s 68.2 Gb/s
Max size 16 GB 16 GB

Multimedia (ISP)

Storage type UFS 4.0, UFS 4.1 UFS 3.1, UFS 4.0
Max display resolution 3840 x 2400 3840 x 2400
Max camera resolution 1x 200MP 1x 200MP
Video capture 8K at 30FPS, 4K at 120FPS 8K at 30FPS, 4K at 120FPS
Video playback 8K at 30FPS, 4K at 120FPS 8K at 30FPS, 4K at 120FPS
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

Connectivity

Modem MediaTek M90 Exynos 5400c
5G support Yes Yes
Download speed Up to 12000 Mbps -
Wi-Fi 7 7
Bluetooth 6.0 6.0
Navigation GPS, GLONASS, Beidou, Galileo, QZSS GPS, GLONASS, Beidou, Galileo, QZSS

Info

Announced August 2026 August 2025
Class Flagship Flagship
Model number GS601 GS501

Cast your vote

So, which SoC would you choose?
148 (87.1%)
22 (12.9%)
Total votes: 170

Related Comparisons

1. Google Tensor G6 vs MediaTek Dimensity 9500s
2. Google Tensor G6 vs Apple A19 Pro
3. Google Tensor G6 vs Apple A19
4. Google Tensor G6 vs Qualcomm Snapdragon 8 Elite Gen 5 V Series
5. Google Tensor G6 vs Qualcomm Snapdragon 8 Elite Gen 5
6. Google Tensor G5 vs Qualcomm Snapdragon 8 Gen 3 (13)
7. Google Tensor G5 vs Apple A17 Pro
8. Google Tensor G5 vs Qualcomm Snapdragon 8 Elite (Gen 4) (21)
9. Google Tensor G5 vs Google Tensor G4 (25)
10. Google Tensor G5 vs MediaTek Dimensity 9400 (3)
Compare other chipsets (190+)

Comments

Please give your opinion on the comparison of Tensor G5 and Tensor G6, or ask any questions
🌐 Register your profile and become part of NanoReview community!
Avatar
Randy chan August 30, 2026 at 04:48 AM
Something important to know, the Tensor G6 is engineeed by TSMC, by contrast the Tensor G5 is engineeed by the inferior Samsung That alone significantly sets those 2 chips apart in favor of the Tensor G6
+4 Reply
Avatar
fuuk1 August 17, 2026 at 09:07 PM
Google should've done 8-core (2+4+2) CPU config this generation instead of 7-core (1+4+2) CPU config. But I guess, the CPU die size might eat into the precious space of the GPU if the CPU has two prime C1-Ultra cores so I guess, it's an okay trade-off. Still :( Oh maybe, they could've used two C1-Premium cores as prime cores, that would've done something. Well, whatever. I just dream.
0 Reply
Avatar
Lexip August 16, 2026 at 02:17 PM
The G6 transistor size is actually 3nm not 2nm as "leaked" before.
+1 Reply
Avatar
Serhii August 17, 2026 at 07:40 AM
Yeah man, it is already fixed :)
0 Reply
Avatar
MSSaar August 15, 2026 at 06:45 AM
G6 is 3 nm, not 2. Please check.
0 Reply
Avatar
Serhii August 17, 2026 at 07:44 AM
Yes, it is. Thank you for a comment.
0 Reply
Avatar
Fatih August 14, 2026 at 01:26 PM
Hey Serhii, Google's VP of Hardware Peng Yu-Chun said G6 is 3nm not 2nm
+1 Reply
Avatar
Serhii August 17, 2026 at 07:53 AM
Yes, specs already updated according to new info.
+2 Reply
Avatar
Toylet August 14, 2026 at 09:04 AM
Its insane how they are still insisting on using a shitty PowerVR GPU for the latest Tensor chip Even an older midrange D8400's GPU vastly outperforms it. At this point, they are better off using Snapdragon or Mediatek SoCs Even if Snapdragon is more expensive, we shouldnt forget that Google burned a lot of money for the R&D development of Tensor that it completely offsets any price advantage tensor might have. Ok some might say, Tensor has better AI. Let me remind you guys, the G5 is only 5-10% better than the 8 Elite in PURE AI tasks. Let me ask you guys, is 10% better AI worth it for VASTLY worse battery life, gaming performance and efficiency? The same thing applies for the G6 vs 9500/8EliteGen5 Tensor is trash.
+8 Reply
Avatar
Mkroni August 14, 2026 at 10:35 AM
We haven't actually seen what the GPU is capable of and geekbench GPU score doesn't translate to gaming performance not at all
0 Reply
Avatar
Phenro August 14, 2026 at 08:37 PM
Mkroni, It is a 5 year old GPU. It would be a miracle if it can even reach the same GPU performance as Snapdragon 8 Gen 1. Also, word of advice, from another tech enthusiast such as me. Please don't defend greedy companies using terrible hardware. You're the one paying for it. Never forget who works for who, and who pays who...
+3 Reply
Avatar
Toylet August 15, 2026 at 02:13 AM
Mkroni, Same argument was made for Tensor G5 and its GPU In the end, it was still trash There is just no hiding the fact that Imagination's PowerVR GPUs are vastly inferior to Mali and Adreno. And we are not even talking about the vastly inferior drivers and developer support that PowerVR has. Its inferior in every way
+1 Reply
Avatar
Samuel28 August 15, 2026 at 08:38 AM
Well its the same thing why apple use powervr gpus before, if they keep updating those graphics, it would outperform the adreno and mali gpus(only if devs arent bored)
0 Reply
Avatar
Toylet August 19, 2026 at 04:29 PM
Samuel28, The last time Apple used PowerVR was 10 years ago, your arguement is already invalid. And now, updating graphics wont make up for actual raw hardware limitations. As Phreno pointed out in the comment above, the underlying architecture of the PowerVR CXTP is 5 years old. So it has inferior drivers, poor developer support, and inferior hardware. No amount of coping that you do will make a G6 Power VR GPU perform better
+4 Reply
Avatar
Ilya August 13, 2026 at 06:40 PM
Both chipsets are suck
+8 Reply
Avatar
LGs August 13, 2026 at 06:02 PM
7 core on G6, no 8 core.
0 Reply
Avatar
Serhii August 14, 2026 at 09:58 AM
Appreciate your useful comment.
0 Reply
EnglishРусский